Pics or it didn't happen – How does social media access affect what we know about killings in the Syrian conflict?
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
| Title |
| |
| Title of Series | ||
| Part Number | 161 | |
| Number of Parts | 177 | |
| Author | ||
| License | CC Attribution - ShareAlike 3.0 Germany: You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor and the work or content is shared also in adapted form only under the conditions of this | |
| Identifiers | 10.5446/31855 (DOI) | |
| Publisher | ||
| Release Date | ||
| Language | ||
| Production Place | Berlin |
Content Metadata
| Subject Area | ||
| Genre | ||
| Abstract |
|
re:publica 2015161 / 177
1
3
4
8
10
12
20
21
23
28
29
30
34
37
40
46
48
49
54
55
59
63
65
67
70
71
72
73
76
86
87
89
91
96
97
98
102
103
110
117
121
127
132
136
142
149
153
157
158
170
171
175
00:00
Right angleMathematical analysisGroup actionVideo gameSpacetimeMeasurementUniform resource locatorMultiplication signType theoryXMLComputer animationLecture/ConferenceMeeting/Interview
00:51
CatastrophismRevision controlEvent horizonHypermediaSmartphoneVideoconferencingComputer-assisted translationFamilyInformationYouTubeComputer animation
01:36
Real-time operating systemRevision controlInformationFigurate numberMetropolitan area networkInternetworkingTraffic reportingOrder (biology)Event horizonHypermediaMathematical analysisGroup actionGoodness of fitCentralizer and normalizerVideoconferencingPrice indexComplete informationComputer animation
03:44
Scaling (geometry)NumberRevision controlOrder (biology)StatisticsMultiplication signLecture/Conference
04:31
MereologyPairwise comparisonStatisticsInformationRevision controlLecture/Conference
05:03
Different (Kate Ryan album)Process (computing)HypermediaRevision controlPhysical systemRow (database)StatisticsInformationFile archiverType theoryTraffic reportingDynamical systemFamilyQuantumContent (media)Right angleLecture/ConferenceComputer animation
06:12
Revision controlArithmetic meanBitHypermediaLecture/ConferenceMeeting/Interview
06:47
Source codeBitSoftware testingRevision controlDifferent (Kate Ryan album)Uniform resource locatorTraffic reportingMereologyLecture/ConferenceComputer animation
08:10
Information securityTraffic reportingSound effectOrder (biology)Multiplication signSource codeAreaRevision controlVideo gameLecture/ConferenceMeeting/Interview
09:04
Point (geometry)InternetworkingDivisorMereologyVideoconferencingCivil engineeringOrder (biology)CASE <Informatik>Different (Kate Ryan album)HypermediaShooting methodMathematicsMetropolitan area networkMultiplication signRevision controlLocal ringTraffic reportingComputer animation
11:09
Different (Kate Ryan album)InternetworkingUniform resource locatorRevision controlMultiplication signCASE <Informatik>HypermediaArmLecture/ConferenceComputer animation
12:16
InternetworkingProcess (computing)Group actionReal-time operating systemSoftwareRight angleInformationAreaStatisticsReal numberComputer animation
13:25
Formal verificationProcess (computing)YouTubeLink (knot theory)InformationDifferent (Kate Ryan album)WebsiteOrder (biology)
13:59
Group actionAreaInformationInternetworkingRevision controlMultiplication signDisk read-and-write headSelf-organizationSoftwareGame controllerRight angleType theorySeries (mathematics)Computer animation
14:57
Level (video gaming)Centralizer and normalizerTime zoneGame controllerInformationAreaGroup actionGreen's functionLecture/Conference
15:38
Figurate numberGroup actionCASE <Informatik>MultiplicationGame controllerRevision controlOrder (biology)Observational studySource codeInformationParallel portComputer animationLecture/Conference
16:18
Source codeComputer clusterInformationType theoryMultiplicationProcess (computing)Electronic mailing listEndliche ModelltheorieGenderTraffic reportingComputer animation
16:58
Type theoryMathematicsEstimatorUniverse (mathematics)Point (geometry)CASE <Informatik>Multiplication signInformationBookmark (World Wide Web)Group actionStatisticsNumberElectronic mailing listWordEntire function1 (number)Figurate numberCircleRight angleEndliche Modelltheorie3 (number)Source codeState observerLecture/ConferenceMeeting/InterviewDiagram
20:28
Table (information)DatabaseSource code1 (number)Traffic reportingComputer animation
21:19
Sound effectCircleSource codeDifferent (Kate Ryan album)NumberElectronic mailing listMappingInformationEstimatorFigurate numberLine (geometry)Price indexRule of inferenceGreatest elementEntire functionTotal S.A.Graph (mathematics)Right angleShape (magazine)Constraint (mathematics)Lecture/ConferenceXML
23:34
Line (geometry)Graph (mathematics)InformationEstimatorMereologyMultiplication signGroup actionNumberLogicCountingFrequencyRow (database)Computer animationDiagram
25:11
Right angleRow (database)Observational studyDirection (geometry)CountingNumberComplete informationInformationHypermediaMultiplication signLevel (video gaming)Line (geometry)CASE <Informatik>Computer animationDiagram
26:59
Group actionCurveEstimatorPattern languageInformationRevision controlEstimationLecture/ConferenceMeeting/Interview
27:55
Computer virusSelf-organizationCASE <Informatik>NumberInformationProcess (computing)Group actionChannel capacitySource codeLecture/ConferenceMeeting/Interview
28:46
Game controllerComplete metric spaceValidity (statistics)MereologySoftware developerLecture/Conference
29:28
Level (video gaming)Formal verificationMultiplicationGroup actionInformationComplete informationMathematical analysisMereologyLimit (category theory)WordComplete metric spaceLecture/ConferenceMeeting/Interview
30:33
Information privacyPerspective (visual)Kernel (computing)Classical physicsSource codeDigitizingCASE <Informatik>Group actionRight angleInformation security1 (number)InformationLecture/ConferenceMeeting/Interview
31:40
Form (programming)Level (video gaming)Traffic reportingGroup actionMathematical analysisInformationComputer virusBitLecture/ConferenceMeeting/Interview
32:27
Expert systemCASE <Informatik>InformationRevision controlContext awarenessCountingNumberRight angleLecture/ConferenceMeeting/Interview
33:57
Revision controlOperator (mathematics)AuthorizationComputer virusLecture/ConferenceMeeting/Interview
34:33
InformationMultiplication signNumberGraph (mathematics)MathematicsLecture/ConferenceMeeting/Interview
35:37
Revision controlObservational studyGraph (mathematics)NumberGraph (mathematics)Right angleLecture/ConferenceMeeting/Interview
36:18
Entire functionSet (mathematics)Category of beingSource codeGroup actionWebsiteCASE <Informatik>Row (database)Projective planeTraffic reportingRevision controlMathematical analysisDifferent (Kate Ryan album)Multiplication signInformationAtomic numberFrequencyTerm (mathematics)Lecture/ConferenceMeeting/Interview
37:49
Source codeRevision controlArithmetic meanLecture/Conference
38:24
Group actionRevision controlTriangulation (psychology)InformationSource codeStatisticsLecture/ConferenceMeeting/Interview
38:55
Level (video gaming)Group actionRight angleMathematical analysisLecture/ConferenceMeeting/Interview
39:27
StatisticsCommitment schemeComputer scienceDecision theoryProcess (computing)Descriptive statisticsPhysical systemRight angleMathematicianUniverse (mathematics)Declarative programmingLecture/ConferenceMeeting/Interview
40:03
Universe (mathematics)Lecture/ConferenceComputer animation
Transcript: English(auto-generated)
00:01
Thank you very much. So my name is Nisha Goudes and I'm a postdoctoral researcher at the University of Mannheim.
00:24
I also work for the Human Rights Data Analysis Group. In the research that I do in my everyday life, I look at how governments use repression, where they use it, what types of repression they use, and how it varies from time to time and from location to location. And in this research, I spend a lot of time thinking about how to accurately measure violence.
00:46
It's surprisingly hard and this is what this talk is going to be about. So the ability to use social media to communicate directly from catastrophes, from conflict
01:01
situations, is something that is well known to people who are attending Republika and basically what we see nowadays is that anyone who's caught in the midst of a violent event can just kind of swipe out their smartphone, take a picture and tweet it for example, or put a video of it on YouTube.
01:22
And what this consequently means is that not only is the way in which we are sharing pictures of cats and pictures of our food and pictures of maybe our babies with our family and our friends, changing the amount of information we create, but it also changes
01:43
the amount of real-time information we have about violence and conflicts. And oftentimes nowadays, we find that we know much more about what's happening in a conflict when we look at social media, for example, than if we look at mainstream media. So this in general is very good news and it's particularly good news when we're talking
02:01
about autocratic countries where the central newspapers, for example, are censored and we have no other way of getting at this information. So in the olden days, it might have taken years and years and years for us to find out about really big massacres and what happened there and nowadays, if one person is present and taking a video of it, we already have an indication of what's going on.
02:24
But what I want to focus on in this talk today is that more information doesn't necessarily mean complete information. And this is something that we tend to forget when we think about the amount of data that we have increasing. So first of all, I'd like to talk to you about why we even document violence, why
02:44
I and my team at the Human Rights Day Analysis Group even pursue this work every day. Then I'd like to talk to you about why some events in conflicts get reported and why some don't and how the internet has changed the way that this reporting happens.
03:04
And in order to figure out how it's changed, we need to know what the dark figure of violence is, what exactly is happening, what the entire reality is. So I'm going to talk to you very simply about how we can figure out what this dark figure
03:22
looks like. And then lastly, we're going to look at how violence is being documented in the Syrian conflict and how much of what's actually going on we're able to see through social media. Okay, so why do we document violence or why do we attempt to document violence and not
03:43
just individual qualitative accounts of what happened to individual people but at a large scale? Well, first of all, we think it's important to remember every single victim in and of itself. But going further than that, at the end of conflicts, oftentimes the question is who was the main perpetrator, who needs to be held accountable, and who can we bring
04:05
to justice? And in order to do that, we need hard statistics. And that is something that drives us. Numbers on conflicts are something that basically shape the post-conflict narrative about what
04:24
happened. So who was the main person in charge of killing the majority of people within a conflict? Who were the main victims within a conflict? And all of these things really are dependent on the information that we gather. And lastly, if we want to have policy-based, a policy of advice that's evidence-based,
04:45
we need some hard facts. We need to know if the Syrian conflict has gotten more intense from 2012 to 2013 or if fewer people have been killed. We need to know which parts of the country are most affected, and so on and so forth. And to answer these questions of comparisons, we need good statistics.
05:04
So we've been doing this job for much longer than social media has existed or my team has been doing this work for much longer. And there are lots of different ways to figure out who's been killed by whom in a conflict.
05:20
One way is to talk to people. Ask them, has your brother been killed? Has your sister been killed? Who in your community has been killed? Who were the perpetrators? How did it happen? Another way is to turn to police records. So what you see here in the middle is a picture of the Guatemalan police archive. And this gives us very clear information of the types of things that the Guatemalan
05:41
police ordered to be done to its citizens. We can turn to truth commission reports. So on the right here, you see this is the truth commission report that was compiled for East Timor. It has an incredible amount of information, qualitative interviews with people, what happened to them, and what the dynamics were within the conflict.
06:03
Another way is to use border records. So how many people have fled a country? How many people have crossed a border? This is the kind of official statistics that we can use to figure out how many people have left a country, how intense the conflict is. And lastly, when we have no other means of figuring out who's been killed, we can
06:24
go and look at graveyards. We can do an actual graveyard census and see who's been killed. Now, most of us, before social media, when we wanted to find out what happened in a conflict, we turned to newspapers. We looked at what newspapers were writing. So a newspaper told us in capital X, 20 people were killed on Sunday, for example.
06:46
And most of us know that this is probably just a bit of what's happening on the ground. Newspapers don't give us a full overview of what's happening. They report to us what they come across. And in the same way, all of these sources have certain shortcomings and only give us
07:05
a certain snapshot of what's happening within a conflict. And there are many different reasons for why these different sources can be biased. The main first one is location. So if we think about newspapers, they might send journalists to a capital city, and the
07:21
journalist will report on what happened in the capital city. Might not necessarily report on something that's happening out in the countryside, right? So this would be already only giving us a part of what's happening. Depending on who the perpetrator is, we might find out about what happened and whatnot. Governments tend to have a lot of resources to cover up what they do.
07:43
Lots of people have disappeared, and there are lots of different ways to cover up a It also depends on the resources within a country. And public health is also something that really affects how much we know about what's happening.
08:03
So if you imagine a country with a very warm climate, a lot of people have been killed, it's going to be top priority to bury those people in the most respectful and quick way in order to avoid any public health risks. This directly affects what we know about what's happening, because by the time reporters
08:21
arrive, people might have already been buried. And lastly, witness incentives are incredibly important. So am I scared if I tell a reporter about what happened that the government will also kill me? And this directly also affects the security situation. So we generally see people reporting from areas that aren't too dangerous to report
08:44
from, because otherwise they'd be risking their own life. So all of these things really affect what we know about what's happening in a conflict and whatnot. And all these things affected our previous sources as well.
09:02
So all data on violence that we have is incomplete, or almost all of it. And the point I want to make with this talk is that large amounts of data don't change this. So just because we have much more now doesn't mean it's still not blending out certain parts of what's happening.
09:20
OK, so counting the dead before the internet. How do we count the dead now that we have the internet? This is a picture of the Syrian conflict. And it's one of those cases where people have been using all kinds of devices to connect and to directly report of what's happening on the ground. And this is absolutely phenomenal, because it really helps us to understand what's happening,
09:44
even if there's no local media that's reporting on it. But at the same time, reporting through the internet is also dependent on certain factors. So I summarized them here as four different factors.
10:03
Documented violence through social media is dependent on people. If no one is present when someone's being killed to film what happened, then we don't know what happened. If you think about a very different case now, recently in the US, there was another police shooting of an unarmed black man,
10:23
and a witness filmed it. Now, if we didn't have that video, then nobody would have known what was happening. And in the same way, in civil war, if there's no one there to take a picture, we're not going to know what's happening. Then the person who's present needs to have a device in order to communicate what he's doing,
10:42
what occurred via social media. The person needs to be able to connect to the internet at some point, and the person needs to be willing to take this picture and upload it. And there might be all kinds of reasons why the person is not willing. It might be political affiliation.
11:01
It might be that the person is way too scared to do this, because he or she might be tracked. It might be because there might be all kinds of different reasons why this might occur. So we see just because we now have the internet to communicate these atrocities directly
11:25
doesn't necessarily mean that it's going to happen every single time. And again, these things are also affected by location and who the perpetrator is, and so on and so forth. Syrian conflict is a case that I'm going to talk about in this talk,
11:44
mostly because it's been termed the most socially mediated conflict in the history of the world. And what we see happening in Syria is likely to be a precursor of what we will see in future conflicts in countries that are also very socially connected.
12:01
Thousands and thousands of people have been killed. The country has been at war for more than four years now, and we see pictures like these every day on social media and in the news. Now, there are also other people using the internet in Syria. One of them is President Assad, and he maintains a pretty lively Instagram account,
12:24
where if you compare this to this picture, Syria looks like a pretty different place. It looks like it's pretty orderly, and everything is fine, and he's doing a good job, and his wife is also taking care of people all over the country. So we also have that picture of Syria.
12:43
But luckily, we have some incredibly dedicated groups working on the ground, providing near real-time information of victims. And I'm just going to mention a few of these here. There are lots more, but some of them are the Syrian Network for Human Rights, the Violations Documentation Center in Syria,
13:01
the Syrian Center for Statistics and Research, the Syrian Observatory for Human Rights, and the Damascus Center for Human Rights Studies. And they're more. And what they're basically doing is they're trying to, in real time, collect as much information as they can on the people who've been killed, how they've been killed, where they were killed, and who's responsible.
13:25
So to give you an example, this is the website. You can go and have a look at it later if you want to, of the Violations Documentation Center. And what they basically do is they have people all over the country gathering as much information as they can on victims.
13:42
And they do this in a very rigorous way. They use different verification processes where they can, they provide a YouTube link of the person who's been killed in order to make sure that the information that they're gathering is correct. Another example is the Syrian Network for Human Rights,
14:02
who also does similar work. And they have thousands and thousands of people who are providing them with information. They're Skyping with them every day. The heads of these organizations aren't even located in Syria anymore. So they're communicating through the internet all the time and collecting all this information. So we really wouldn't have as much information as we have now
14:23
without the internet in the Syrian conflict. And the work that these groups are doing is absolutely phenomenal. It's phenomenal not because it's exciting or crazy, but because it's absolutely courageous what they're doing. So I encourage you to check out what their work looks like.
14:42
But one common feature or one of the problems that all of them communicate to us is that they have a lot of problems getting information in areas that are besieged by the government. So where the government has moved in and taken over entire control. Because people are fearful of telling them what's happening, and because the government is basically hindering their work.
15:05
Now this is a map of Syria from August 2013 that was created by the Carter Center. And what it shows us is the central zones of territorial control within the country. And what you can see in the south is areas that were predominantly
15:23
government controlled in red in August 2013. Green was predominantly opposition controlled, and blue was predominantly Kurdish held. Now if these groups are telling us that the information they have on what is government controlled is far less than the information that they have on where the opposition is controlled,
15:44
this means that we know much less about what's happening in the south than what's happening in the north. So this directly affects what we know about the conflict, and if we only look at observed data now, we might come to very wrong conclusions about who the main perpetrator is in the conflict.
16:07
So in order to figure out how much we actually know, we need to estimate the dark figure. And luckily we have all these groups in Syria working in parallel on the same cases. So we have multiple data sources, and what we can do is we can compare the information between the different sources.
16:24
Because all of these sources are telling us the name of the person who was killed, where the person was killed, possibly the gender, and the date. So this can help us to compare the information. And then these overlaps can be used to model the reporting process.
16:42
So we can try and understand who was included on a list and who wasn't. And once we have this model, we can statistically predict or estimate what went unreported. Now I'm going to walk you very briefly through the type of research that we do.
17:02
And I'm going to use no math, but I'm happy to talk about the math afterwards. OK, so let's assume that we're trying to estimate how many people were killed in this village at a certain point in time. And these are all the people, this is the universe, these are all the people who were killed.
17:21
So the first group goes in and tries to figure out who was killed. And they document these people. This is the first group. And the second group goes in and tries to find out again all the people who were killed. They're able to document these people. Now already we know that the three people in the middle
17:41
between these two circles have been documented on both lists. Whereas the people on the outside haven't. And then another group comes in. And the third group documents these people. And again here we can look at the overlaps. We can see how many people have been included on three lists,
18:00
how many have been included on these two, on those two, and then on the individual ones. So this is what we mean when we talk about the overlap of the different sources. And as you can see, these ones on the side are the dark figure. No one has actually included those on a list. Now the great thing about statistics is, using the probability of these overlaps,
18:21
we can estimate how big this number is. So depending on the size of the overlaps, we can extrapolate what the size of the dark figure is. So to give you an example, this might be three sources that we have, and this might be the entire universe that we're trying to estimate.
18:44
Now we might have a different case where the groups go in, they collect much more information, but the universe, the dark figure, is just the same. So if we compare this case and this case, in the second case, we've documented far more people,
19:03
so it looks like more is happening, but the dark figure is the same. So what this means is these groups, as you can see, have basically collected the same information. They have a few people that the other groups haven't collected.
19:21
So the size of the overlap is what's really important, not so much the overall size of the observed. And then this might be another case where we have very few people who've been reported, but also the overlaps are very small, right? So we know that probably what's out there is much more,
19:41
and we can estimate that. Now I'm happy to talk about the statistical models behind this afterwards. It's surprisingly hard to do this work. But if you're interested in the nitty-gritty details, I'm only too happy to talk about them.
20:01
So let's apply this to the Syrian case. What we did for the UN last year was establish exactly the amount of people who had been reported in the different lists. Because the UN wasn't interested so much
20:20
in understanding the entire amount of people who've been killed but the number of people who'd been documented. So that's what we did. And newspapers rightly reported on the fact that we counted 191,000 people, but the truth is probably much, much worse.
20:45
So to give you an idea of the data that we're looking at now, this is the only table I'm going to show, by the way. This is a snapshot of the database, or what the database looks like. We have a name for every single victim, and then we have multiple sources. So here you have source A, B, C, D.
21:04
And then the ones and the zeros tell us whether that individual victim was included in source A, B, C, or D. So as you can see here, the first victim listed was only included in source A. He, we see as sexist male,
21:22
was killed in Aleppo on the 16th of November 2012. The second person that we recorded was found in two different lists, in list C and D, killed in Homs in 2013, and was also male, and so on and so forth. So this is the information that we then use.
21:42
And this perfectly maps to the different circles that you just saw. Another way to look at this information is to think about how many people were included in more than one source. So what you see here is the amount of killings that were recorded,
22:01
that occurred in Syria, by different governorates. So right at the top, most people were recorded to have been killed in rural Damascus. And the darker the shade of the bar is, the more sources include that victim's name. And the lighter the shade is, the less sources include that victim's name.
22:22
Now if you think back to the circles, we know that if the overlaps are quite small, the total number is likely to be pretty large. So already from this graph, we can see that the number of people who were only included in one source are a lot, which means that the dark figure is likely to be really, really large.
22:42
So this already offers us a first indication of the fact that we have a lot of data on what's happening in Syria. We have almost 200,000 people on the list, and still it's only showing us a snapshot of what's happening.
23:04
So what we see here is the estimate. This is actually looking at the entire dark figure, and this is looking at the entire dark figure for the rural areas surrounding rural Damascus, from May 2013 to May 2014.
23:20
And the little lines that you can see at the bottom, the little black lines, are the actual sources that we included. And the yellow line is when we put them all together. And the blue line gives us the estimated total. Now as you can see, the blue line is much, much higher than the other two lines.
23:41
And you'll probably notice that there's a really big spike in mid-August in 2013. And for those of you who have been following the news more closely, you know that in August 2013,
24:00
the Syrian chemical weapon attack occurred. And the Washington Post reported back then that 104,000 people were killed in this attack. Now if we look at this graph, we obviously see that this is an undercount. And it's not surprising, because when so many people have been killed, who's supposed to take up all that information? Who's supposed to record all that information at once?
24:22
It's basically impossible. It's going to take us years to find out all the names of these people. So this graph can help us understand where the observed information is leaving out important parts of the story.
24:41
Now in this graph, we're only looking at people who were killed by bombings, people who were killed by shellings, and then big, large attacks. So it makes sense that this would produce a really big undercount at the time when the chemical attack occurred. Now it might be that these groups in two years have way more names,
25:01
and that then the observed number approximates what we've estimated. And to show you for the same time period, these are killings that occurred in a targeted way. So people who were first detained and then killed, people who were tortured and then killed,
25:21
people who were executed directly. And as you can see, during the chemical attack, not a lot of people were directly detained. So in this case, the observed count actually gives us a pretty good idea of what happened. But then later in November,
25:41
in the outskirts of Damascus, intense battles again were going on between the Syrian government and the opposition. And here we see that there was a really big spike at the end of November and the beginning of December where far more arrests and targeted executions occurred than the data actually tells us.
26:01
So later when we try to figure out who was responsible for massacres, who was responsible for certain atrocities, it's going to be important to think about not only what we can, about the observed data that we have, but also what the actual number might look like.
26:21
So in conclusion, what I'd like to leave you with is that big data, because it's this big buzzword that everyone's talking about now, big data on human rights can really create an illusion of complete information. And it's so tempting to take this information generated through social media and just map it, put it on a map, for example,
26:41
or put it on a timeline and say this is what's happening in Syria. Because it's really easy to do that. It's already digitized, right? If you had to do that with all the police records in Guatemala, it would take you much, much longer. And this illusion is dangerous when it allows us to draw false conclusions
27:01
about patterns within the conflict. And importantly, this is not a criticism of the data, so without these groups' information, we wouldn't even be able to do the estimates. It's the criticism of drawing inaccurate conclusions. And thank you, and I'm open to questions.
27:27
Thank you very much. Are there any questions? Or annotations, or something like that?
27:48
Hello, my name is Daniel. I have just one question. It sounds like you only assume overestimation of death. Is it also a possibility of an underestimation?
28:00
I mean, I would definitely underestimate them, and they could also be overestimated by the organizations because they have a political interest. I don't know whether that's the case, but I'm just wondering. That's a really good question. So one step in the processing of the data that I left out is that we take the information from these groups,
28:21
and we deduplicate it. So we go through and do internal matching first. And that is because most groups don't have the capacity to do that. They'll get information about a person who was killed from two different sources, and then, again, if we used the big data as it was there, we might actually overestimate exactly. So the first step in actually arriving at a total observed number
28:43
is doing deduplication. So basically the data you've shown is from 2011, and you said that the data from governmental-controlled Syria
29:00
is not so valid like the data from the oppositional Syria, but now ISIS is playing a bigger role in Syria. Are there any conclusions about how valid is the data coming from these parts of the country right now?
29:21
So I think we need to distinguish between data validity and data completeness. Data validity is the question of, is the data that we get correct or not? And as I said, most of these groups go through multiple stages of verification, and they include pictures of the people who've been killed. They include all kinds of biographical information
29:43
that makes it quite convincing that these are people who've, you know, this is accurate information. A different question is the level of completeness, and yes, these groups have reported to us that as soon as ISIS has taken over a certain part of the country or as soon as the government has taken over a certain part of the country,
30:01
the amount of information goes down, or it's more hard for them to access information. So we don't include inaccurate information in this analysis. If we don't have complete information on a person who's been killed, so if we don't have a name, then we can't include that person in the analysis.
30:29
Hi, I just have a quick question. Would you be able to comment on the limitations to anonymization using kind of classic tactics of giving numeric codes to names,
30:44
particularly from a digital privacy and data protection perspective? So do you mean in the Syrian case or in general? So we need full identifying information to do our work,
31:03
and groups that share their data privately with us do it through a secure way. The human rights groups working with us are the ones who decide how they transmit their information. We communicate to them, look, we need this information to do our work, and we're happy to exchange it with you through encrypted channels.
31:22
We sign non-disclosure agreements with you, but actually a lot of these groups put their data online. And the reason for this is that these are people who've been killed. So the security risk for the victim itself is not there anymore.
31:41
It's very different when we're talking about non-lethal forms of violence. So if we're talking about sexual violence, for example, where the level of reporting is much lower even, or if we're talking about other forms of violence that are non-lethal, anonymization is a much bigger concern. But generally we let the groups dictate how they share their information with us,
32:03
and we don't make the names publicly available when we do the analysis. Hello. Could you please elaborate? You mentioned a teeny bit about it, but what advice, concrete advice do you have for journalists in a reporting situation?
32:23
And maybe the Syrian example, but also in general, what we can do to take care with these numbers? That's a great question. I think infographics are incredibly...
32:43
There's something that people like to look at, right? Infographics are something that get a lot of clicks. And in a lot of cases, when we know that this is the full information included in the infographic, then it's fine to go ahead with that. I think the main recommendation for journalists when it comes to data on violence
33:01
is to really think about where that information comes from, to talk to experts and get kind of an opinion on whether it's useful in this case to visualize the information in a certain way. I think that's the concrete advice that I'd give. It's really dependent on context as well.
33:24
So for example, there are certain conflicts where we have full information. We now know almost every single person who was killed in Kosovo. So this is a case where I'd say, you know, go ahead, make an infographic, do it.
33:43
But there are other cases where we know that the data that's lying underneath it is very problematic, both because there are duplicates in it and because it undercounts. I don't know if that's satisfactory or not. One more question from me. First of all, thank you for your work.
34:01
I think it's very important what you do and your colleagues. The question is, from a statistical viewpoint, is violence decreasing or not? Because we're now shocked by 300,000 people, 300,000 people who died in the Syrian conflict, but we're looking back at World War II and other wars,
34:23
and I think Steven Pinker was an author who argued that we're actually facing less violence. On the other hand, we are much more vulnerable towards violence because we are not so used to it. From the statistical and your scientific viewpoint,
34:41
so said 200,000 dead people may be, I could imagine that if they were really serious about killing in Syria, they could have killed even more. So on your question to Syria, I think we know that the true number is higher.
35:00
And I think every single person who's been killed in Syria is a tragedy, just to make that clear. But I think the number is probably much higher than 200,000, which these graphs also show. So trusting a set number that's out there is problematic. And from a scientific viewpoint and from the research I do,
35:21
my honest opinion is that we don't have the information to infer whether the world is becoming more peaceful or not. Because the information we have has changed so rapidly across time. And the question is, is it important or not?
35:41
But I think I'll leave that to more philosophical people. The work that we do is try to, as objectively as possible, come up with answers for concrete conflicts. What we do know is if we look at the observed data that we have, there's a massive spike now through Syria again in the past few years.
36:02
And even there we know that it's an undercount. So I think that the graphs that I show here are quite clear in that they show us that the number is much higher than 200,000. How do you decide which groups you're working with?
36:23
Like when you take another case, an Assyrian case, where it seems like you have really trustful sources, how do you find these groups and how do you begin to trust in their sources and what they say? So on our website we have an entire set of categories or questions that we go through
36:45
when we think about whether we want to take on a new project or not. And one of the things is do we think that our work will have an impact? So there are certain conflicts where we think the truth is not going to make a difference. So finding out what's happening might actually not make a difference at all.
37:02
And those aren't necessarily cases where we're going to spend all our manpower on, because we're very few people working on this. In terms of sources, so in the Syrian conflict, these groups were put in contact with us through the UN primarily. And I should say that in the report,
37:22
we also include information that was gathered by the Syrian government. The Syrian government gave us information for the first three months of the conflict and that was to 2,000 records. So that's also included in the analysis for the UN report.
37:40
It's not included in the analysis that I showed you earlier on, because that was for an earlier time period. It's really interesting what you said. I was going to ask because it's so much work. I realised suddenly how important and how many people have to help you. So how can you decide what you just said now?
38:00
We have to choose which conflicts we are going to try to get the data. Does it depend on who can give you the means, the financial means, to have people working on it? And you said sometimes it's not worth finding the truth. So what is your criteria that you can do all the hard work to get the data?
38:23
So the first criteria is we need multiple sources to do our work. Where there's only one group collecting information, we can't do any triangulations. It's still incredibly important work these groups are doing within the conflict. It's just that we can't use them for our statistics. So we need more than just one or two sources that we can work with.
38:45
And then as you saw, we need really high quality information. It's not just enough that we can read somewhere five people were killed on day X. We actually need to know the names of those people and so on and so forth. And I think that level of quality really also ensures that the groups that we're working with
39:01
are fully dedicated to what they're doing. It's maybe a more slightly personal question, but I think you're doing some fascinating work. And I'm just wondering, are you driven more by the data,
39:20
the analysis of the data, or by the human rights aspect of things? What drives you? So my background is political science. I'm a political scientist turned quantitative political scientist turned maybe applied statistics user. So my background is human rights, definitely.
39:40
And I think all of the people who work for us and we have pure mathematicians, we have statisticians working for us, we have computer scientists working for us. One of the things that we say in our job descriptions is we need a full commitment to the universal declaration of human rights. So all of us are intrinsically driven, I'd say. Yes.
40:02
Any questions? Okay, Anita. Thank you very much for the insights and your work. Thank you.