[Script Info]
Title: 
[Events]
Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text
Dialogue: 0,0:00:10.08,0:00:17.98,Default,,0000,0000,0000,,{\i1}applause{\i0}
Dialogue: 0,0:00:17.98,0:00:22.90,Default,,0000,0000,0000,,Thank you very much, can you…\NYou can hear me? Yes!
Dialogue: 0,0:00:22.90,0:00:27.62,Default,,0000,0000,0000,,I’ve been at this now 23 years. We\Nworked, with… My colleagues and I,
Dialogue: 0,0:00:27.62,0:00:31.39,Default,,0000,0000,0000,,we worked in about 30 countries,\Nwe’ve advised 9 Truth Commissions,
Dialogue: 0,0:00:31.39,0:00:36.41,Default,,0000,0000,0000,,official Truth Commissions, 4 UN missions,
Dialogue: 0,0:00:36.41,0:00:40.15,Default,,0000,0000,0000,,4 international criminal tribunals.\NWe have testified in 4 different cases
Dialogue: 0,0:00:40.15,0:00:44.24,Default,,0000,0000,0000,,– 2 internationally, 2 domestically – and\Nwe’ve advised dozens and dozens
Dialogue: 0,0:00:44.24,0:00:49.12,Default,,0000,0000,0000,,of non-governmental Human Rights groups\Naround the world. The point of this stuff
Dialogue: 0,0:00:49.12,0:00:54.18,Default,,0000,0000,0000,,is to figure out how to bring the\Nknowledge of the people who’ve suffered
Dialogue: 0,0:00:54.18,0:00:58.77,Default,,0000,0000,0000,,human rights violations to bear,\Non demanding accountability
Dialogue: 0,0:00:58.77,0:01:04.96,Default,,0000,0000,0000,,from the perpetrators. Our job is to\Nfigure out how we can tell the truth.
Dialogue: 0,0:01:04.96,0:01:09.24,Default,,0000,0000,0000,,It is one of the moral foundations of the\Ninternational Human Rights movement
Dialogue: 0,0:01:09.24,0:01:14.22,Default,,0000,0000,0000,,that we speak Truth to Power. We\Nlook in the face of the powerful
Dialogue: 0,0:01:14.22,0:01:19.30,Default,,0000,0000,0000,,and we tell them what we believe\Nthey have done that is wrong.
Dialogue: 0,0:01:19.30,0:01:23.64,Default,,0000,0000,0000,,If that’s gonna work, we\Nhave to speak the truth.
Dialogue: 0,0:01:23.64,0:01:29.47,Default,,0000,0000,0000,,We have to be right, we\Nhave to get the analysis on.
Dialogue: 0,0:01:29.47,0:01:33.98,Default,,0000,0000,0000,,That’s not always easy and to get there,
Dialogue: 0,0:01:33.98,0:01:37.21,Default,,0000,0000,0000,,there are sort of 3 themes that\NI wanna try to touch in this talk.
Dialogue: 0,0:01:37.21,0:01:40.38,Default,,0000,0000,0000,,Since the talk is pretty short I’m\Nreally gonna touch on 2 of them, so
Dialogue: 0,0:01:40.38,0:01:43.62,Default,,0000,0000,0000,,at the very end of the talk I’ll invite\Npeople who’d like to talk more about
Dialogue: 0,0:01:43.62,0:01:49.27,Default,,0000,0000,0000,,the specifically technical aspects of this\Nwork, about classifiers, about clustering,
Dialogue: 0,0:01:49.27,0:01:53.62,Default,,0000,0000,0000,,about statistical estimation, about\Ndatabase techniques. People who wanna talk
Dialogue: 0,0:01:53.62,0:01:56.99,Default,,0000,0000,0000,,about that I’d love to gather and we’ll\Ntry to find a space. I’ve been fighting
Dialogue: 0,0:01:56.99,0:02:00.46,Default,,0000,0000,0000,,with the Wiki for 2 days; I think\NI’m probably not the only one.
Dialogue: 0,0:02:00.46,0:02:04.96,Default,,0000,0000,0000,,We can gather, we can talk about\Nthat stuff more in detail. So today,
Dialogue: 0,0:02:04.96,0:02:09.99,Default,,0000,0000,0000,,in the next 25 minutes I’m\Ngoing to focus specifically on
Dialogue: 0,0:02:09.99,0:02:14.52,Default,,0000,0000,0000,,the trial of General\NJosé Efraín Ríos Montt
Dialogue: 0,0:02:14.52,0:02:20.20,Default,,0000,0000,0000,,who ruled Guatemala from\NMarch 1982 until August 1983.
Dialogue: 0,0:02:20.20,0:02:25.18,Default,,0000,0000,0000,,That’s General Ríos, there in\Nthe upper corner in the red tie.
Dialogue: 0,0:02:25.18,0:02:30.60,Default,,0000,0000,0000,,During the government\Nof General Ríos Montt
Dialogue: 0,0:02:30.60,0:02:35.61,Default,,0000,0000,0000,,tens of thousands of people were killed by\Nthe army of Guatemala. And the question
Dialogue: 0,0:02:35.61,0:02:39.61,Default,,0000,0000,0000,,that has been facing Guatemalans\Nsince that time is:
Dialogue: 0,0:02:39.61,0:02:44.08,Default,,0000,0000,0000,,“Did the pattern of killing\Nthat the army committed
Dialogue: 0,0:02:44.08,0:02:49.69,Default,,0000,0000,0000,,constitute acts of genocide?”. Now\Ngenocide is a very specific crime
Dialogue: 0,0:02:49.69,0:02:54.42,Default,,0000,0000,0000,,in International Law. It does not\Nmean you killed a lot of people.
Dialogue: 0,0:02:54.42,0:02:58.91,Default,,0000,0000,0000,,There are other war crimes for mass\Nkilling. Genocide specifically means
Dialogue: 0,0:02:58.91,0:03:03.93,Default,,0000,0000,0000,,that you picked out a particular group;\Nand to the exclusion of other groups
Dialogue: 0,0:03:03.93,0:03:08.46,Default,,0000,0000,0000,,nearby them you focused\Non eliminating that group.
Dialogue: 0,0:03:08.46,0:03:14.24,Default,,0000,0000,0000,,That’s key because for a statistician\Nthat gives us a hypothesis we can test
Dialogue: 0,0:03:14.24,0:03:18.86,Default,,0000,0000,0000,,which is: “What is the relative risk,\Nwhat is the differential probability
Dialogue: 0,0:03:18.86,0:03:22.82,Default,,0000,0000,0000,,of people in the target group being\Nkilled relative to their neighbours
Dialogue: 0,0:03:22.82,0:03:28.15,Default,,0000,0000,0000,,who are not in the target group?”\NSo without further ado,
Dialogue: 0,0:03:28.15,0:03:31.97,Default,,0000,0000,0000,,let’s look at the relative risk of\Nbeing killed for indigenous people
Dialogue: 0,0:03:31.97,0:03:36.88,Default,,0000,0000,0000,,in the 3 rural counties of\NChajul, Cotzal and Nebaj
Dialogue: 0,0:03:36.88,0:03:41.40,Default,,0000,0000,0000,,relative to their\Nnon-indigenous neighbours.
Dialogue: 0,0:03:41.40,0:03:45.96,Default,,0000,0000,0000,,We have – and I’ll talk in a moment about\Nhow we have this – we have information,
Dialogue: 0,0:03:45.96,0:03:51.49,Default,,0000,0000,0000,,and evidence, and estimations of the\Ndeaths of about 2150 indigenous people.
Dialogue: 0,0:03:51.49,0:03:58.55,Default,,0000,0000,0000,,People killed by the army in the period\Nof the government of General Ríos.
Dialogue: 0,0:03:58.55,0:04:02.55,Default,,0000,0000,0000,,The population, the total number of\Npeople alive who were indigenous
Dialogue: 0,0:04:02.55,0:04:07.37,Default,,0000,0000,0000,,in those counties in the census\Nof 1981 is about 39,000.
Dialogue: 0,0:04:07.37,0:04:14.50,Default,,0000,0000,0000,,So the approximate crude mortality\Nrate due to homicide by the army
Dialogue: 0,0:04:14.50,0:04:18.71,Default,,0000,0000,0000,,is 5.5% for indigenous people in\Nthat period. Now that’s relative
Dialogue: 0,0:04:18.71,0:04:22.89,Default,,0000,0000,0000,,to the homicide rate for non-indigenous\Npeople in the same place
Dialogue: 0,0:04:22.89,0:04:27.20,Default,,0000,0000,0000,,of approximately 0.7%. So what\Nwe ask is: “What is the ratio
Dialogue: 0,0:04:27.20,0:04:30.53,Default,,0000,0000,0000,,between those 2 numbers?” And\Nthe ratio between those 2 numbers
Dialogue: 0,0:04:30.53,0:04:35.60,Default,,0000,0000,0000,,is the relative risk. It’s approximately\N8. We interpret that as: if you were
Dialogue: 0,0:04:35.60,0:04:41.34,Default,,0000,0000,0000,,an indigenous person alive in\None of those 3 counties in 1982,
Dialogue: 0,0:04:41.34,0:04:46.94,Default,,0000,0000,0000,,your probability of being killed\Nby the army was 8 times greater
Dialogue: 0,0:04:46.94,0:04:51.07,Default,,0000,0000,0000,,than a person also living\Nin those 3 counties
Dialogue: 0,0:04:51.07,0:04:56.18,Default,,0000,0000,0000,,who was not indigenous.\NEight times, 8 times!
Dialogue: 0,0:04:56.18,0:05:00.25,Default,,0000,0000,0000,,To put that in relative terms: the\Nprobability… the relative risk of being
Dialogue: 0,0:05:00.25,0:05:04.72,Default,,0000,0000,0000,,a Bosniac relative to being Serb\Nin Bosnia during the war in Bosnia
Dialogue: 0,0:05:04.72,0:05:09.80,Default,,0000,0000,0000,,was a little less than 3. So your\Nrelative risk of being indigenous
Dialogue: 0,0:05:09.80,0:05:13.31,Default,,0000,0000,0000,,was more than twice nearly 3 times\Nas much as your relative risk
Dialogue: 0,0:05:13.31,0:05:19.20,Default,,0000,0000,0000,,of being Bosniac in the Bosnian War.\NIt’s an astonishing level of focus.
Dialogue: 0,0:05:19.20,0:05:23.81,Default,,0000,0000,0000,,It shows a tremendous planning\Nand coherence, I believe.
Dialogue: 0,0:05:23.81,0:05:29.47,Default,,0000,0000,0000,,So, again coming back to the statistical\Nconclusion, how do we come to that?
Dialogue: 0,0:05:29.47,0:05:32.85,Default,,0000,0000,0000,,How do we find that information? How do we\Nmake that conclusion? First, we’re only
Dialogue: 0,0:05:32.85,0:05:35.47,Default,,0000,0000,0000,,looking at homicides committed by the\Narmy. We’re not looking at homicides
Dialogue: 0,0:05:35.47,0:05:39.41,Default,,0000,0000,0000,,committed by other parties, by\Nthe guerrillas, by private actors.
Dialogue: 0,0:05:39.41,0:05:44.50,Default,,0000,0000,0000,,We’re not looking at excess mortality,\Nthe mortality that we might find
Dialogue: 0,0:05:44.50,0:05:47.71,Default,,0000,0000,0000,,in conflict that is in excess of\Nnormal peacetime mortality.
Dialogue: 0,0:05:47.71,0:05:51.47,Default,,0000,0000,0000,,We’re not looking at any of that,\Nonly homicide. And the percentage
Dialogue: 0,0:05:51.47,0:05:55.33,Default,,0000,0000,0000,,relates the number of people killed by the\Narmy with the population that was alive.
Dialogue: 0,0:05:55.33,0:05:58.65,Default,,0000,0000,0000,,That’s crucial here. We’re looking at\Nrates and we’re comparing the rate
Dialogue: 0,0:05:58.65,0:06:02.43,Default,,0000,0000,0000,,of the indigenous people shown in the\Nblue bar to non-indigenous people
Dialogue: 0,0:06:02.43,0:06:06.87,Default,,0000,0000,0000,,shown in the green bar. The width of\Nthe bars show the relative populations
Dialogue: 0,0:06:06.87,0:06:11.83,Default,,0000,0000,0000,,in each of those 2 communities. So clearly\Nthere are many more indigenous people,
Dialogue: 0,0:06:11.83,0:06:14.98,Default,,0000,0000,0000,,but a higher fraction of them are also\Nkilled. The bars also show something else.
Dialogue: 0,0:06:14.98,0:06:18.05,Default,,0000,0000,0000,,And that’s what I’ll focus on for the\Nrest of the talk. There are 2 sections
Dialogue: 0,0:06:18.05,0:06:22.16,Default,,0000,0000,0000,,to each of the 2 bars, a dark section\Non the bottom, a lighter section on top.
Dialogue: 0,0:06:22.16,0:06:27.78,Default,,0000,0000,0000,,And what that indicates is what we know\Nin terms of being able to name people
Dialogue: 0,0:06:27.78,0:06:31.25,Default,,0000,0000,0000,,with their first and last name, their\Nlocation and dates of death, and
Dialogue: 0,0:06:31.25,0:06:35.56,Default,,0000,0000,0000,,what we must infer statistically. Now I’m\Nbeginning to touch on the second theme
Dialogue: 0,0:06:35.56,0:06:40.95,Default,,0000,0000,0000,,of my talk: Which is that when we are\Nstudying mass violence and war crimes,
Dialogue: 0,0:06:40.95,0:06:48.75,Default,,0000,0000,0000,,we cannot do statistical or pattern\Nanalysis with raw information.
Dialogue: 0,0:06:48.75,0:06:51.95,Default,,0000,0000,0000,,We must use the tools of mathematical\Nstatistics to understand
Dialogue: 0,0:06:51.95,0:06:56.08,Default,,0000,0000,0000,,what we don’t know! The information\Nwhich cannot be observed directly.
Dialogue: 0,0:06:56.08,0:07:00.65,Default,,0000,0000,0000,,We have to estimate that in order to\Ncontrol for the process of the production
Dialogue: 0,0:07:00.65,0:07:04.99,Default,,0000,0000,0000,,of information. Information doesn’t just\Nfall out of the sky, the way it does
Dialogue: 0,0:07:04.99,0:07:10.36,Default,,0000,0000,0000,,for industry. If I’m running an ISP I know\Nevery packet that runs through my routers.
Dialogue: 0,0:07:10.36,0:07:14.96,Default,,0000,0000,0000,,That’s not how the social world works. In\Norder to find information about killings
Dialogue: 0,0:07:14.96,0:07:17.89,Default,,0000,0000,0000,,we have to hear about that killing from\Nsomeone, we have to investigate,
Dialogue: 0,0:07:17.89,0:07:22.12,Default,,0000,0000,0000,,we have to find the human remains.\NAnd if we can’t observe the killing
Dialogue: 0,0:07:22.12,0:07:28.13,Default,,0000,0000,0000,,we won’t hear about it and many killings\Nare hidden. In my team we have a kind of
Dialogue: 0,0:07:28.13,0:07:33.76,Default,,0000,0000,0000,,catch phrase: that the world… if a lawyer\Nis killed in a big city at high noon
Dialogue: 0,0:07:33.76,0:07:38.26,Default,,0000,0000,0000,,the world knows about it before\Ndinner time. Every single time.
Dialogue: 0,0:07:38.26,0:07:41.85,Default,,0000,0000,0000,,But when a rural peasant is killed 3-days\Nwalk from a road in the dead of night,
Dialogue: 0,0:07:41.85,0:07:45.49,Default,,0000,0000,0000,,we’re unlikely to ever hear. And\Ntechnology is not changing this.
Dialogue: 0,0:07:45.49,0:07:48.90,Default,,0000,0000,0000,,I’ll talk later about that technology is\Nactually making the problem worse.
Dialogue: 0,0:07:48.90,0:07:53.47,Default,,0000,0000,0000,,So, let’s get back to Guatemala\Nand just conclude
Dialogue: 0,0:07:53.47,0:07:57.95,Default,,0000,0000,0000,,that the little vertical bars, little\Nvertical lines at the top of each bar
Dialogue: 0,0:07:57.95,0:08:03.08,Default,,0000,0000,0000,,indicate the confidence interval. Which is\Nsimilar to what lay people sometimes call
Dialogue: 0,0:08:03.08,0:08:07.20,Default,,0000,0000,0000,,a margin of error. It is our level of\Nuncertainty about each of those estimates
Dialogue: 0,0:08:07.20,0:08:10.96,Default,,0000,0000,0000,,and you’ll notice that the uncertainty\Nis much, much smaller than
Dialogue: 0,0:08:10.96,0:08:14.51,Default,,0000,0000,0000,,the difference between the 2 bars. The\Nuncertainty does not affect our ability
Dialogue: 0,0:08:14.51,0:08:17.97,Default,,0000,0000,0000,,to draw the conclusion that there\Nwas a spectacular difference
Dialogue: 0,0:08:17.97,0:08:21.90,Default,,0000,0000,0000,,in the mortality rates between the\Npeople who were the hypothesized
Dialogue: 0,0:08:21.90,0:08:26.63,Default,,0000,0000,0000,,target of genocide and those who were not.
Dialogue: 0,0:08:26.63,0:08:30.52,Default,,0000,0000,0000,,Now the data: first we\Nhad the census of 1981,
Dialogue: 0,0:08:30.52,0:08:35.34,Default,,0000,0000,0000,,this was a crucial piece. I think there’s\Nvery interesting questions to ask
Dialogue: 0,0:08:35.34,0:08:39.61,Default,,0000,0000,0000,,about why the Government of Guatemala\Nconducted a census on the eve of
Dialogue: 0,0:08:39.61,0:08:44.54,Default,,0000,0000,0000,,committing a genocide. There is excellent\Nwork done by historical demographers
Dialogue: 0,0:08:44.54,0:08:47.95,Default,,0000,0000,0000,,about the use of censuses in mass\Nviolence. It has been common
Dialogue: 0,0:08:47.95,0:08:52.88,Default,,0000,0000,0000,,throughout history. Similarly,\Nor excuse me, in parallel
Dialogue: 0,0:08:52.88,0:08:57.42,Default,,0000,0000,0000,,there were 4 very large\Nprojects. First, the CIIDH
Dialogue: 0,0:08:57.42,0:09:01.60,Default,,0000,0000,0000,,– a group of non-Governmental\NHuman Rights groups –
Dialogue: 0,0:09:01.60,0:09:06.61,Default,,0000,0000,0000,,collected 1240 records of deaths\Nin this three-county region.
Dialogue: 0,0:09:06.61,0:09:11.75,Default,,0000,0000,0000,,Next, the Catholic Church collected\Na bit fewer than 800 deaths.
Dialogue: 0,0:09:11.75,0:09:16.54,Default,,0000,0000,0000,,The truth commission – the Comisión\Npara el Esclarecimiento Histórico (CEH) –
Dialogue: 0,0:09:16.54,0:09:22.00,Default,,0000,0000,0000,,conducted a really big research\Nproject in the late 1990s and
Dialogue: 0,0:09:22.00,0:09:25.81,Default,,0000,0000,0000,,of that we got information about a little\Nbit more than a thousand deaths.
Dialogue: 0,0:09:25.81,0:09:30.45,Default,,0000,0000,0000,,And then the National Program for\NCompensation is very, very large
Dialogue: 0,0:09:30.45,0:09:35.37,Default,,0000,0000,0000,,and gave us about 4700\Nrecords of deaths.
Dialogue: 0,0:09:35.37,0:09:40.66,Default,,0000,0000,0000,,Now, this is interesting\Nbut this is not unique.
Dialogue: 0,0:09:40.66,0:09:45.77,Default,,0000,0000,0000,,Many of the deaths are reported in common\Nacross those data sources and so…
Dialogue: 0,0:09:45.77,0:09:49.49,Default,,0000,0000,0000,,we think about this in terms of a Venn\Ndiagram. We think of: how did these
Dialogue: 0,0:09:49.49,0:09:54.33,Default,,0000,0000,0000,,different data sets intersect with each\Nother or collide with each other. And
Dialogue: 0,0:09:54.33,0:09:59.13,Default,,0000,0000,0000,,we can diagram that as in the sense\Nof these 3 white circles intersecting.
Dialogue: 0,0:09:59.13,0:10:05.61,Default,,0000,0000,0000,,But as I mentioned earlier we’re also\Ninterested in what we have not observed.
Dialogue: 0,0:10:05.61,0:10:09.49,Default,,0000,0000,0000,,And this is crucial for us because\Nwhen we’re thinking about
Dialogue: 0,0:10:09.49,0:10:13.42,Default,,0000,0000,0000,,how much information we have, we have to\Ndistinguish between the world on the left,
Dialogue: 0,0:10:13.42,0:10:17.20,Default,,0000,0000,0000,,in which our intersecting circles\Ncover about a third of the reality,
Dialogue: 0,0:10:17.20,0:10:21.83,Default,,0000,0000,0000,,versus the world on the right where our\Nintersecting circles cover all of reality.
Dialogue: 0,0:10:21.83,0:10:26.39,Default,,0000,0000,0000,,These are very different worlds; and the\Nreason they’re so different is not simply
Dialogue: 0,0:10:26.39,0:10:29.71,Default,,0000,0000,0000,,because we want to know the magnitude,\Nnot simply because we want to know
Dialogue: 0,0:10:29.71,0:10:34.49,Default,,0000,0000,0000,,the total number of killings. That’s\Nimportant – but even more important:
Dialogue: 0,0:10:34.49,0:10:40.16,Default,,0000,0000,0000,,we have to know that we’ve covered,\Nwe’ve estimated in equal proportions
Dialogue: 0,0:10:40.16,0:10:44.43,Default,,0000,0000,0000,,the two parties. We have to estimate in\Nequal proportions the number of deaths
Dialogue: 0,0:10:44.43,0:10:48.34,Default,,0000,0000,0000,,of non-indigenous people and the\Nnumber of deaths of indigenous people.
Dialogue: 0,0:10:48.34,0:10:51.51,Default,,0000,0000,0000,,Because if we don’t get those\Nestimates correct our comparison
Dialogue: 0,0:10:51.51,0:10:56.08,Default,,0000,0000,0000,,of their mortality rates will be biased.\NOur story will be wrong. We will fail
Dialogue: 0,0:10:56.08,0:11:01.84,Default,,0000,0000,0000,,to speak Truth to Power. We can’t have\Nthat. So what do we do? Algebra!
Dialogue: 0,0:11:01.84,0:11:06.39,Default,,0000,0000,0000,,Algebra is our friend. So I’m gonna\Ngive you just a tiny taste of how we
Dialogue: 0,0:11:06.39,0:11:09.65,Default,,0000,0000,0000,,solve this problem and I’m going to\Nintroduce a series of assumptions.
Dialogue: 0,0:11:09.65,0:11:13.28,Default,,0000,0000,0000,,Those of you who would like to debate\Nthose assumptions: I invite you to join me
Dialogue: 0,0:11:13.28,0:11:18.36,Default,,0000,0000,0000,,after the talk and we will talk endlessly\Nand tediously about capture heterogeneity.
Dialogue: 0,0:11:18.36,0:11:22.24,Default,,0000,0000,0000,,But in the short term,
Dialogue: 0,0:11:22.24,0:11:27.94,Default,,0000,0000,0000,,we have a universe N of total killings in\Na specific time/space/ethnicity/location.
Dialogue: 0,0:11:27.94,0:11:30.69,Default,,0000,0000,0000,,And of that we have 2 projects A and B.
Dialogue: 0,0:11:30.69,0:11:34.62,Default,,0000,0000,0000,,A captures some number of\Ndeaths from the universe N,
Dialogue: 0,0:11:34.62,0:11:40.17,Default,,0000,0000,0000,,and the probability with which a death is\Ncaptured by project A from the universe N
Dialogue: 0,0:11:40.17,0:11:44.60,Default,,0000,0000,0000,,is by elementary probability theory the\Nnumber of deaths documented by A
Dialogue: 0,0:11:44.60,0:11:48.74,Default,,0000,0000,0000,,divided by the unknown number\Nof deaths in the population N.
Dialogue: 0,0:11:48.74,0:11:52.97,Default,,0000,0000,0000,,Similarly, the probability with which a\Ndeath from N is documented by project B
Dialogue: 0,0:11:52.97,0:11:58.15,Default,,0000,0000,0000,,is B over N, and this is the cool part:\Nthe probability with which a death
Dialogue: 0,0:11:58.15,0:12:01.95,Default,,0000,0000,0000,,is documented by both A and B is M.
Dialogue: 0,0:12:01.95,0:12:05.58,Default,,0000,0000,0000,,Now we can put the 2 databases together,\Nwe can compare them. Let’s talk about
Dialogue: 0,0:12:05.58,0:12:09.37,Default,,0000,0000,0000,,the use of random force classifiers\Nand clustering to do that later.
Dialogue: 0,0:12:09.37,0:12:12.49,Default,,0000,0000,0000,,But we can put the 2 databases together,\Ncompare them, determine the deaths
Dialogue: 0,0:12:12.49,0:12:17.43,Default,,0000,0000,0000,,that are in M – that is in N both\NA and B – and divide M by N.
Dialogue: 0,0:12:17.43,0:12:23.06,Default,,0000,0000,0000,,But, also by probability theory, the\Nprobability that a death occurs in M
Dialogue: 0,0:12:23.06,0:12:27.74,Default,,0000,0000,0000,,is equal to the product of\Nthe individual probabilities.
Dialogue: 0,0:12:27.74,0:12:31.62,Default,,0000,0000,0000,,The probability of any compound event, an\Nevent made up of two independent events is
Dialogue: 0,0:12:31.62,0:12:36.41,Default,,0000,0000,0000,,equal to the product of those two\Nevents, so M over N is equal to
Dialogue: 0,0:12:36.41,0:12:41.42,Default,,0000,0000,0000,,A over N times B over N. Solve for N.
Dialogue: 0,0:12:41.42,0:12:45.14,Default,,0000,0000,0000,,Multiply it through by N squared, divide\Nby M, and we have an estimate of N
Dialogue: 0,0:12:45.14,0:12:49.36,Default,,0000,0000,0000,,which is equal to AB over M. Now, the\Nlights in my eyes, I can’t see, but I saw
Dialogue: 0,0:12:49.36,0:12:52.74,Default,,0000,0000,0000,,a few light bulbs go off over people’s\Nheads. And when I showed this proof
Dialogue: 0,0:12:52.74,0:12:57.18,Default,,0000,0000,0000,,to the judge in the trial of General Ríos
Dialogue: 0,0:12:57.18,0:13:01.53,Default,,0000,0000,0000,,I saw a light bulb go on over her head.
Dialogue: 0,0:13:01.53,0:13:04.38,Default,,0000,0000,0000,,It’s a beautiful thing,\Nit’s a beautiful thing.
Dialogue: 0,0:13:04.38,0:13:09.51,Default,,0000,0000,0000,,{\i1}applause{\i0}
Dialogue: 0,0:13:09.51,0:13:12.66,Default,,0000,0000,0000,,So we don’t do it in 2 systems because\Nthat takes a lot of assumptions.
Dialogue: 0,0:13:12.66,0:13:16.07,Default,,0000,0000,0000,,We do it in 4. You will recall that we\Nhave 4 data sources. We organize
Dialogue: 0,0:13:16.07,0:13:21.53,Default,,0000,0000,0000,,the data sources in this format\Nsuch that we have an inclusion
Dialogue: 0,0:13:21.53,0:13:26.25,Default,,0000,0000,0000,,and an exclusion pattern in the table on \Nthe left, which… for which we can define
Dialogue: 0,0:13:26.25,0:13:29.81,Default,,0000,0000,0000,,the number of deaths which fall into\Neach of these intersecting patterns.
Dialogue: 0,0:13:29.81,0:13:33.73,Default,,0000,0000,0000,,And I’ll give you a very quick\Nmetaphor here. The metaphor is:
Dialogue: 0,0:13:33.73,0:13:38.24,Default,,0000,0000,0000,,imagine that you have 2 dark rooms and you\Nwant to assess the size of those 2 rooms
Dialogue: 0,0:13:38.24,0:13:42.05,Default,,0000,0000,0000,,– which room is larger? And the only\Ntool that you have to assess the size
Dialogue: 0,0:13:42.05,0:13:46.36,Default,,0000,0000,0000,,of those rooms is a handful of little\Nrubber balls. The little rubber balls
Dialogue: 0,0:13:46.36,0:13:50.40,Default,,0000,0000,0000,,have a property that when they hit each\Nother they make a sound. {\i1}makes CLICK sound{\i0}
Dialogue: 0,0:13:50.40,0:13:53.39,Default,,0000,0000,0000,,So we throw the balls into the first\Nroom and we listen, and we hear
Dialogue: 0,0:13:53.39,0:13:57.19,Default,,0000,0000,0000,,{\i1}makes several CLICK sounds{\i0}. We\Ncollect the balls, go to the second room,
Dialogue: 0,0:13:57.19,0:14:00.49,Default,,0000,0000,0000,,throw them with equal force – imagining\Na spherical cow of uniform density!
Dialogue: 0,0:14:00.49,0:14:03.95,Default,,0000,0000,0000,,We throw the balls into the second\Nroom with equal force and we hear
Dialogue: 0,0:14:03.95,0:14:07.80,Default,,0000,0000,0000,,{\i1}makes one CLICK sound{\i0}\NSo which room is larger?
Dialogue: 0,0:14:07.80,0:14:12.07,Default,,0000,0000,0000,,The second room, because we hear fewer\Ncollisions, right? Well, the estimation,
Dialogue: 0,0:14:12.07,0:14:15.62,Default,,0000,0000,0000,,the toy example I gave in the previous\Nslide is the mathematical formalization
Dialogue: 0,0:14:15.62,0:14:20.07,Default,,0000,0000,0000,,of the intuition that fewer\Ncollisions mean a larger space.
Dialogue: 0,0:14:20.07,0:14:23.33,Default,,0000,0000,0000,,And so what we’re doing here is\Nlaying out the pattern of collisions.
Dialogue: 0,0:14:23.33,0:14:26.68,Default,,0000,0000,0000,,Not just the collisions, the pairwise\Ncollisions, but the three-way and
Dialogue: 0,0:14:26.68,0:14:31.41,Default,,0000,0000,0000,,four-way collisions. And that\Nallows us to make the estimate
Dialogue: 0,0:14:31.41,0:14:37.44,Default,,0000,0000,0000,,that was shown in the bar graph of\Nthe light part of each of the bars. So
Dialogue: 0,0:14:37.44,0:14:41.46,Default,,0000,0000,0000,,we can come back to our conclusion and put\Na confidence interval on the estimates.
Dialogue: 0,0:14:41.46,0:14:45.91,Default,,0000,0000,0000,,And the confidence intervals are shown\Nthere. Now I’m gonna move through this
Dialogue: 0,0:14:45.91,0:14:50.85,Default,,0000,0000,0000,,somewhat more quickly to get to the end of\Nthe talk but I wanna put up one more slide
Dialogue: 0,0:14:50.85,0:14:56.24,Default,,0000,0000,0000,,that was used in the testimony\Nand that is that we divided time
Dialogue: 0,0:14:56.24,0:15:01.22,Default,,0000,0000,0000,,into 16-month periods and\Ncompared the 16-month period of
Dialogue: 0,0:15:01.22,0:15:04.58,Default,,0000,0000,0000,,General Ríos’s governance – now it’s only\N16 months ’cause we went April to July,
Dialogue: 0,0:15:04.58,0:15:07.68,Default,,0000,0000,0000,,because it’s only a few days in August, a\Nfew days in March, so we shaved those off,
Dialogue: 0,0:15:07.68,0:15:12.31,Default,,0000,0000,0000,,okay… – 16-month period of General\NRíos’s Government and compared it
Dialogue: 0,0:15:12.31,0:15:17.11,Default,,0000,0000,0000,,to several periods before and after. And\NI think that the key observation here
Dialogue: 0,0:15:17.11,0:15:21.81,Default,,0000,0000,0000,,is that the rate of killing\Nagainst indigenous people
Dialogue: 0,0:15:21.81,0:15:26.73,Default,,0000,0000,0000,,is substantially higher done under General\NRíos’s Government than under previous
Dialogue: 0,0:15:26.73,0:15:33.28,Default,,0000,0000,0000,,or succeeding governments. But more\Nimportantly the ratio between the two,
Dialogue: 0,0:15:33.28,0:15:37.95,Default,,0000,0000,0000,,the relative risk of being killed as an\Nindigenous person, was at its peak
Dialogue: 0,0:15:37.95,0:15:42.64,Default,,0000,0000,0000,,during the government of General Ríos.
Dialogue: 0,0:15:42.64,0:15:46.71,Default,,0000,0000,0000,,Have we proven genocide? No.
Dialogue: 0,0:15:46.71,0:15:49.87,Default,,0000,0000,0000,,This is evidence consistent with the\Nhypothesis that acts of genocide
Dialogue: 0,0:15:49.87,0:15:53.54,Default,,0000,0000,0000,,were committed. The finding of genocide\Nis a legal finding, not so much
Dialogue: 0,0:15:53.54,0:15:58.58,Default,,0000,0000,0000,,a scientific one. So as scientists,\Nour job is to provide evidence that
Dialogue: 0,0:15:58.58,0:16:02.87,Default,,0000,0000,0000,,the finders of fact – the judges in this\Ncase – can use in their determination.
Dialogue: 0,0:16:02.87,0:16:05.22,Default,,0000,0000,0000,,This is evidence consistent\Nwith that hypothesis.
Dialogue: 0,0:16:05.22,0:16:08.19,Default,,0000,0000,0000,,Were this evidence otherwise, as\Nscientists we would say we would
Dialogue: 0,0:16:08.19,0:16:11.48,Default,,0000,0000,0000,,reject the hypothesis that genocide was\Ncommitted. However, with this evidence
Dialogue: 0,0:16:11.48,0:16:15.37,Default,,0000,0000,0000,,we find that the evidence,\Nthe data is consistent with
Dialogue: 0,0:16:15.37,0:16:18.08,Default,,0000,0000,0000,,the prosecution’s hypothesis.
Dialogue: 0,0:16:18.08,0:16:25.32,Default,,0000,0000,0000,,So, it worked!
Dialogue: 0,0:16:25.32,0:16:29.05,Default,,0000,0000,0000,,Ríos Montt was convicted on\Ngenocide charges. {\i1}applause{\i0}
Dialogue: 0,0:16:29.05,0:16:31.36,Default,,0000,0000,0000,,You can clap!\N{\i1}applause{\i0}
Dialogue: 0,0:16:31.36,0:16:36.36,Default,,0000,0000,0000,,{\i1}applause{\i0}
Dialogue: 0,0:16:36.36,0:16:39.50,Default,,0000,0000,0000,,For a week!\N{\i1}mumbled, surprised laughter{\i0}
Dialogue: 0,0:16:39.50,0:16:42.28,Default,,0000,0000,0000,,Then the Constitutional Court intervened,
Dialogue: 0,0:16:42.28,0:16:44.96,Default,,0000,0000,0000,,there I know a couple of experts on\NGuatemala here in the audience
Dialogue: 0,0:16:44.96,0:16:47.84,Default,,0000,0000,0000,,who can tell you more about why that\Nhappened and exactly what happened.
Dialogue: 0,0:16:47.84,0:16:52.67,Default,,0000,0000,0000,,However, the Constitutional\NCourt ordered a new trial,
Dialogue: 0,0:16:52.67,0:16:59.16,Default,,0000,0000,0000,,which is at this time scheduled\Nfor the very beginning of 2015.
Dialogue: 0,0:16:59.16,0:17:02.97,Default,,0000,0000,0000,,And I look forward to testifying again,
Dialogue: 0,0:17:02.97,0:17:06.82,Default,,0000,0000,0000,,and again, and again, and again!
Dialogue: 0,0:17:06.82,0:17:12.68,Default,,0000,0000,0000,,{\i1}applause{\i0}
Dialogue: 0,0:17:12.68,0:17:16.99,Default,,0000,0000,0000,,Look, but I wanna come back to this point.\NBecause as a bunch of technologists…
Dialogue: 0,0:17:16.99,0:17:21.59,Default,,0000,0000,0000,,– there is a lot of folks who really like\Ntechnology here, I really like it too!
Dialogue: 0,0:17:21.59,0:17:25.56,Default,,0000,0000,0000,,Technology doesn’t get us to science\N– you have to have science
Dialogue: 0,0:17:25.56,0:17:28.77,Default,,0000,0000,0000,,to get you to science. Technology helps\Nyou organize the data. It helps you do
Dialogue: 0,0:17:28.77,0:17:32.05,Default,,0000,0000,0000,,all kinds of extremely great and cool\Nthings without which we wouldn’t be able
Dialogue: 0,0:17:32.05,0:17:36.48,Default,,0000,0000,0000,,to even do the science. But you\Ncan’t have just technology!
Dialogue: 0,0:17:36.48,0:17:40.97,Default,,0000,0000,0000,,You can’t just have a bunch of data\Nand make conclusions. That’s naive,
Dialogue: 0,0:17:40.97,0:17:44.53,Default,,0000,0000,0000,,and you will get the wrong conclusions.\N‘The point of rigorous statistics is
Dialogue: 0,0:17:44.53,0:17:48.10,Default,,0000,0000,0000,,to be right’, and there is a little bit of\Na caveat there – or to at least know
Dialogue: 0,0:17:48.10,0:17:51.62,Default,,0000,0000,0000,,how uncertain you are. Statistics is often\Ncalled the ‘Science of Uncertainty’.
Dialogue: 0,0:17:51.62,0:17:55.96,Default,,0000,0000,0000,,That is actually my favorite\Ndefinition of it. So,
Dialogue: 0,0:17:55.96,0:18:01.51,Default,,0000,0000,0000,,I’m going to assume that we\Ncare about getting it right.
Dialogue: 0,0:18:01.51,0:18:05.49,Default,,0000,0000,0000,,No one laughed, that’s good.
Dialogue: 0,0:18:05.49,0:18:08.89,Default,,0000,0000,0000,,Not everyone does, to my distress.
Dialogue: 0,0:18:08.89,0:18:11.32,Default,,0000,0000,0000,,So if you only have some of the data
Dialogue: 0,0:18:11.32,0:18:15.49,Default,,0000,0000,0000,,– and I will argue that we always\Nonly have some of the data –
Dialogue: 0,0:18:15.49,0:18:20.45,Default,,0000,0000,0000,,you need some kind of model that will tell\Nyou the relationship between your data
Dialogue: 0,0:18:20.45,0:18:23.99,Default,,0000,0000,0000,,and the real world.\NStatisticians call that an inference.
Dialogue: 0,0:18:23.99,0:18:26.20,Default,,0000,0000,0000,,In order to get from here to there\Nyou’re gonna need some kind of
Dialogue: 0,0:18:26.20,0:18:30.47,Default,,0000,0000,0000,,probability model that tells you\Nwhy your data is like the world,
Dialogue: 0,0:18:30.47,0:18:33.96,Default,,0000,0000,0000,,or in what sense you have to tweet,\Ntwiddle and do algebra with your data
Dialogue: 0,0:18:33.96,0:18:39.31,Default,,0000,0000,0000,,to get from what you can\Nobserve to what is actually true.
Dialogue: 0,0:18:39.31,0:18:42.69,Default,,0000,0000,0000,,And statistics is about comparisons.\NYeah, we get a big number and
Dialogue: 0,0:18:42.69,0:18:46.17,Default,,0000,0000,0000,,journalists love the big number; but\Nit’s really about these relationships
Dialogue: 0,0:18:46.17,0:18:50.61,Default,,0000,0000,0000,,and patterns! So to get those\Nrelationships and patterns,
Dialogue: 0,0:18:50.61,0:18:53.56,Default,,0000,0000,0000,,in order for them to be right, in order\Nfor our answer to be correct,
Dialogue: 0,0:18:53.56,0:18:57.44,Default,,0000,0000,0000,,every one of the estimates we make\Nfor every point in the pattern
Dialogue: 0,0:18:57.44,0:19:01.70,Default,,0000,0000,0000,,has to be right. It’s a hard\Nproblem. It’s a hard problem.
Dialogue: 0,0:19:01.70,0:19:05.07,Default,,0000,0000,0000,,And what I worry about is that\Nwe have come into this world
Dialogue: 0,0:19:05.07,0:19:09.40,Default,,0000,0000,0000,,in which people throw the notion of Big\NData around as though the data allows us
Dialogue: 0,0:19:09.40,0:19:14.23,Default,,0000,0000,0000,,to make an end-run around problems\Nof sampling and modeling. It doesn’t.
Dialogue: 0,0:19:14.23,0:19:19.12,Default,,0000,0000,0000,,So as technologist, the reason I’m,\Nyou know, ranting at you guys about it
Dialogue: 0,0:19:19.12,0:19:24.54,Default,,0000,0000,0000,,is that it’s very tempting to have a lot\Nof data and think you have an answer!
Dialogue: 0,0:19:24.54,0:19:30.58,Default,,0000,0000,0000,,And it’s even more tempting because\Nin industry context you might be right.
Dialogue: 0,0:19:30.58,0:19:36.74,Default,,0000,0000,0000,,Not so much in Human Rights, not so\Nmuch. Violence is a hidden process.
Dialogue: 0,0:19:36.74,0:19:39.96,Default,,0000,0000,0000,,The people who commit violence have\Nan enormous commitment to hiding it,
Dialogue: 0,0:19:39.96,0:19:44.42,Default,,0000,0000,0000,,distorting it, explaining it in different\Nways. All of those things dramatically
Dialogue: 0,0:19:44.42,0:19:48.35,Default,,0000,0000,0000,,affect the information that is produced\Nfrom the violence that we’re going to use
Dialogue: 0,0:19:48.35,0:19:53.73,Default,,0000,0000,0000,,to do our analysis. So we usually\Ndon’t know what we don’t know
Dialogue: 0,0:19:53.73,0:19:58.22,Default,,0000,0000,0000,,in Human Rights data collection.\NAnd that means that we don’t know
Dialogue: 0,0:19:58.22,0:20:03.83,Default,,0000,0000,0000,,if what we don’t know is systematically\Ndifferent from what we do know.
Dialogue: 0,0:20:03.83,0:20:06.27,Default,,0000,0000,0000,,Maybe we know about all the lawyers\Nand we don’t know about the people
Dialogue: 0,0:20:06.27,0:20:10.07,Default,,0000,0000,0000,,in the countryside. Maybe we know\Nabout all the indigenous people
Dialogue: 0,0:20:10.07,0:20:14.13,Default,,0000,0000,0000,,and not the non-indigenous people.\NIf that were true, the argument
Dialogue: 0,0:20:14.13,0:20:17.98,Default,,0000,0000,0000,,that I just made would be merely\Nan artifact of the reporting process
Dialogue: 0,0:20:17.98,0:20:21.74,Default,,0000,0000,0000,,rather than some true analysis. Now\Nwe did the estimations why I believe
Dialogue: 0,0:20:21.74,0:20:25.01,Default,,0000,0000,0000,,we can reject that critique, but that’s\Nwhat we have to worry about.
Dialogue: 0,0:20:25.01,0:20:28.86,Default,,0000,0000,0000,,And let’s go back to the Venn diagram\Nand say: which of these is accurate?
Dialogue: 0,0:20:28.86,0:20:32.84,Default,,0000,0000,0000,,It’s not just for one of the\Npoints in our pattern analysis.
Dialogue: 0,0:20:32.84,0:20:36.50,Default,,0000,0000,0000,,The problem is that we’re\Ngoing to compare things.
Dialogue: 0,0:20:36.50,0:20:40.89,Default,,0000,0000,0000,,As in Peru where we compared killings\Ncommitted by the Peruvian army against
Dialogue: 0,0:20:40.89,0:20:44.86,Default,,0000,0000,0000,,killings committed by the Maoist Guerillas\Nwith the Sendero Luminoso. And we found
Dialogue: 0,0:20:44.86,0:20:51.46,Default,,0000,0000,0000,,there that in fact we knew very little\Nabout what the Sendero Luminoso had done.
Dialogue: 0,0:20:51.46,0:20:55.78,Default,,0000,0000,0000,,Whereas we knew almost everything\Nwhat the Peruvian army had done.
Dialogue: 0,0:20:55.78,0:20:57.97,Default,,0000,0000,0000,,This is called the coverage rate.\NThe rate between what we know and
Dialogue: 0,0:20:57.97,0:21:02.75,Default,,0000,0000,0000,,what we don’t know. And\Nraw data, however big,
Dialogue: 0,0:21:02.75,0:21:07.51,Default,,0000,0000,0000,,does not get us to patterns.\NAnd here is a bunch of…
Dialogue: 0,0:21:07.51,0:21:11.57,Default,,0000,0000,0000,,kinds of raw data that I’ve used\Nand that I really enjoy using.
Dialogue: 0,0:21:11.57,0:21:14.27,Default,,0000,0000,0000,,You know – truth commission testimonies,\NUN investigations, press articles,
Dialogue: 0,0:21:14.27,0:21:18.31,Default,,0000,0000,0000,,SMS messages, crowdsourcing, NGO\Ndocumentation, social media feeds,
Dialogue: 0,0:21:18.31,0:21:21.18,Default,,0000,0000,0000,,perpetrator records, government archives,\Nstate agency registries – I know those
Dialogue: 0,0:21:21.18,0:21:23.57,Default,,0000,0000,0000,,sound all the same but they actually\Nturn out to be slightly different.
Dialogue: 0,0:21:23.57,0:21:28.34,Default,,0000,0000,0000,,Happy to talk in tedious detail! Refugee\NCamp records, any non-random sample.
Dialogue: 0,0:21:28.34,0:21:31.99,Default,,0000,0000,0000,,All of those are gonna take\Nsome kind of probability model
Dialogue: 0,0:21:31.99,0:21:36.07,Default,,0000,0000,0000,,and we don’t have that many\Nprobability models to use. So
Dialogue: 0,0:21:36.07,0:21:40.33,Default,,0000,0000,0000,,raw data is great for cases – but\Nit doesn’t get you to patterns.
Dialogue: 0,0:21:40.33,0:21:45.12,Default,,0000,0000,0000,,And patterns – again – patterns are\Nthe thing that allow us to do analysis.
Dialogue: 0,0:21:45.12,0:21:49.29,Default,,0000,0000,0000,,They are the thing… the patterns are what\Nget us to something that we can use
Dialogue: 0,0:21:49.29,0:21:53.63,Default,,0000,0000,0000,,to help prosecutors, advocates and the…
Dialogue: 0,0:21:53.63,0:21:56.41,Default,,0000,0000,0000,,and the victims themselves.
Dialogue: 0,0:21:56.41,0:22:00.59,Default,,0000,0000,0000,,I gave a version of this talk, a\Nmuch earlier version of this talk
Dialogue: 0,0:22:00.59,0:22:04.63,Default,,0000,0000,0000,,several years ago in Medellín, Columbia.\NI’ve worked a lot in Columbia,
Dialogue: 0,0:22:04.63,0:22:07.67,Default,,0000,0000,0000,,it’s really… it’s a great place to\Nwork. There’s really terrific
Dialogue: 0,0:22:07.67,0:22:13.57,Default,,0000,0000,0000,,Victims Rights groups there.\NAnd a woman from a township,
Dialogue: 0,0:22:13.57,0:22:17.31,Default,,0000,0000,0000,,smaller than a county, near to Medellín\Ncame up to me after the talk and she said:
Dialogue: 0,0:22:17.31,0:22:21.15,Default,,0000,0000,0000,,“You know, a lot of people… you\Nknow I’m a Human Rights activist,
Dialogue: 0,0:22:21.15,0:22:25.31,Default,,0000,0000,0000,,my job is to collect data, I tell stories\Nabout people who have suffered.
Dialogue: 0,0:22:25.31,0:22:28.21,Default,,0000,0000,0000,,But there are people in my\Nvillage I know who have had
Dialogue: 0,0:22:28.21,0:22:32.91,Default,,0000,0000,0000,,people in their families disappeared and\Nthey’re never gonna talk about, ever.
Dialogue: 0,0:22:32.91,0:22:38.09,Default,,0000,0000,0000,,We’re never going to be able to use\Ntheir names, because they are afraid.”
Dialogue: 0,0:22:38.09,0:22:45.35,Default,,0000,0000,0000,,We can’t name the victims. At\Nleast we’d better count them.
Dialogue: 0,0:22:45.35,0:22:49.52,Default,,0000,0000,0000,,So about that counting: there’s\N3 ways to do it right. You can have
Dialogue: 0,0:22:49.52,0:22:54.43,Default,,0000,0000,0000,,a perfect census – you can have all the\Ndata. Yeah it’s nice, good luck with that.
Dialogue: 0,0:22:54.43,0:22:58.91,Default,,0000,0000,0000,,You can have a random sample\Nof the population - that’s hard!
Dialogue: 0,0:22:58.91,0:23:03.03,Default,,0000,0000,0000,,Sometimes doable but very hard.\NIn my experience we rarely interview
Dialogue: 0,0:23:03.03,0:23:07.14,Default,,0000,0000,0000,,victims of homicide, very rarely.\N{\i1}Laughing{\i0}
Dialogue: 0,0:23:07.14,0:23:09.64,Default,,0000,0000,0000,,And that means there’s a complicated\Nprobability relationship between
Dialogue: 0,0:23:09.64,0:23:13.67,Default,,0000,0000,0000,,the person you sampled, the interview\Nand the death that they talk to you about.
Dialogue: 0,0:23:13.67,0:23:17.30,Default,,0000,0000,0000,,Or you can do some kind of posterior\Nmodeling of the sampling process which is…
Dialogue: 0,0:23:17.30,0:23:21.26,Default,,0000,0000,0000,,which is in essence what\NI proposed in the earlier slide.
Dialogue: 0,0:23:21.26,0:23:25.02,Default,,0000,0000,0000,,So what can we do with raw data,\Nguys? We can collect a bunch of…
Dialogue: 0,0:23:25.02,0:23:28.93,Default,,0000,0000,0000,,We can say that a case exists. Ok\N– that’s actually important! We can say:
Dialogue: 0,0:23:28.93,0:23:34.41,Default,,0000,0000,0000,,“Something happened” with raw data. We can\Nsay: “We know something about that case".
Dialogue: 0,0:23:34.41,0:23:38.25,Default,,0000,0000,0000,,We can say: “There were 100 victims\Nin that case or at least 100 victims
Dialogue: 0,0:23:38.25,0:23:41.57,Default,,0000,0000,0000,,in that case”, if we can name 100 people.
Dialogue: 0,0:23:41.57,0:23:46.39,Default,,0000,0000,0000,,But we can’t do comparisons: “This\Nis the biggest massacre this year”.
Dialogue: 0,0:23:46.39,0:23:48.35,Default,,0000,0000,0000,,We don’t really know. Because we\Ndon’t know about that massacres
Dialogue: 0,0:23:48.35,0:23:53.91,Default,,0000,0000,0000,,we don’t know about. No patterns. Don’t\Ntalk about the hot spot of violence.
Dialogue: 0,0:23:53.91,0:23:59.42,Default,,0000,0000,0000,,No, we don’t know that. Happy to talk\Nmore about that if we gather after,
Dialogue: 0,0:23:59.42,0:24:06.44,Default,,0000,0000,0000,,but I wanna come to a close here with\Nthe importance of getting it right.
Dialogue: 0,0:24:06.44,0:24:11.38,Default,,0000,0000,0000,,I’ve talked about one case today. This\Nis another case, the case of this man:
Dialogue: 0,0:24:11.38,0:24:16.32,Default,,0000,0000,0000,,Edgar Fernando García. Mr. García was\Na student Labor leader in Guatemala
Dialogue: 0,0:24:16.32,0:24:19.80,Default,,0000,0000,0000,,early in the 1980s. He left\Nhis office in February 1984
Dialogue: 0,0:24:19.80,0:24:24.47,Default,,0000,0000,0000,,– did not come home. People reported\Nlater that they saw someone
Dialogue: 0,0:24:24.47,0:24:28.81,Default,,0000,0000,0000,,shoving Mr. García into a\Nvehicle and driving away.
Dialogue: 0,0:24:28.81,0:24:33.90,Default,,0000,0000,0000,,His widow became a very important\NHuman Rights activist in Guatemala
Dialogue: 0,0:24:33.90,0:24:38.57,Default,,0000,0000,0000,,and now she’s a very important, and\Nin my opinion impressive politician.
Dialogue: 0,0:24:38.57,0:24:42.24,Default,,0000,0000,0000,,And there’s her infant daughter. She\Ncontinued to struggle to find out
Dialogue: 0,0:24:42.24,0:24:46.13,Default,,0000,0000,0000,,what had happened to\NMr. García for decades.
Dialogue: 0,0:24:46.13,0:24:50.40,Default,,0000,0000,0000,,And in 2006 documents came to light\Nin the National Archives of the…
Dialogue: 0,0:24:50.40,0:24:54.43,Default,,0000,0000,0000,,excuse me, the Historical Archives\Nof the national Police, showing that
Dialogue: 0,0:24:54.43,0:24:59.32,Default,,0000,0000,0000,,the Police had realized an operation\Nin the area of Mr. García’s office
Dialogue: 0,0:24:59.32,0:25:01.93,Default,,0000,0000,0000,,and it was very likely that\Nthey had disappeared him.
Dialogue: 0,0:25:01.93,0:25:07.40,Default,,0000,0000,0000,,These 2 guys up here in the upper\Nright were Police officers in that area;
Dialogue: 0,0:25:07.40,0:25:11.36,Default,,0000,0000,0000,,they were arrested, charged with the\Ndisappearance of Mister García and
Dialogue: 0,0:25:11.36,0:25:15.62,Default,,0000,0000,0000,,convicted. Part of the evidence used to\Nconvict them was communications meta data
Dialogue: 0,0:25:15.62,0:25:19.51,Default,,0000,0000,0000,,showing that documents\Nflowed through the archive.
Dialogue: 0,0:25:19.51,0:25:23.70,Default,,0000,0000,0000,,I mean paper communications! We coded\Nit by hand. We went through and read
Dialogue: 0,0:25:23.70,0:25:28.46,Default,,0000,0000,0000,,the ‘From’ and ‘To’ lines\Nfrom every Memo. And
Dialogue: 0,0:25:28.46,0:25:34.23,Default,,0000,0000,0000,,they were convicted in 2010\Nand after that conviction
Dialogue: 0,0:25:34.23,0:25:38.70,Default,,0000,0000,0000,,Mr. García’s infant daughter – now\Na grown woman – was clearly joyful.
Dialogue: 0,0:25:38.70,0:25:42.73,Default,,0000,0000,0000,,Justice brings closure to a family\Nthat never knows when to start talking
Dialogue: 0,0:25:42.73,0:25:48.06,Default,,0000,0000,0000,,about someone in the past tense.\NPerhaps even more powerfully:
Dialogue: 0,0:25:48.06,0:25:52.32,Default,,0000,0000,0000,,those guys’ grand boss, their boss's\Nboss, Colonel Héctor Bol de la Cruz,
Dialogue: 0,0:25:52.32,0:25:58.44,Default,,0000,0000,0000,,this man here, was convicted\Nof Mr. García’s disappearance
Dialogue: 0,0:25:58.44,0:26:02.07,Default,,0000,0000,0000,,in September this year [2013].\N{\i1}applause{\i0}
Dialogue: 0,0:26:02.07,0:26:07.61,Default,,0000,0000,0000,,{\i1}applause{\i0}
Dialogue: 0,0:26:07.61,0:26:10.79,Default,,0000,0000,0000,,I don’t know if any of you have\Never been dissident students,
Dialogue: 0,0:26:10.79,0:26:15.33,Default,,0000,0000,0000,,but if you’ve been dissident students\Ndemonstrating in the street think about
Dialogue: 0,0:26:15.33,0:26:19.30,Default,,0000,0000,0000,,how you would feel if your friends\Nand comrades were disappeared,
Dialogue: 0,0:26:19.30,0:26:23.42,Default,,0000,0000,0000,,and take a long look at Colonel Bol\Nde la Cruz. Here is the rest of the stuff
Dialogue: 0,0:26:23.42,0:26:25.63,Default,,0000,0000,0000,,that we will talk about if we gather\Nafterwards. Thank you very much
Dialogue: 0,0:26:25.63,0:26:29.09,Default,,0000,0000,0000,,for your attention. I really\Nhave enjoyed CCC.
Dialogue: 0,0:26:29.09,0:26:36.09,Default,,0000,0000,0000,,{\i1}applause{\i0}
Dialogue: 0,0:26:36.09,0:26:47.20,Default,,0000,0000,0000,,{\i1}Subtitles created by c3subtitles.de\Nin the year 2016. Join and help us!{\i0}\N