快猫短视频

Patrick Ball took his BA in sociology from Columbia in 1988, and his PhD in sociology from the University of Michigan, Ann Arbor, in 1998 (鈥渋t took 10 years because I was never there鈥�). The 25-year-old Science and Human Rights Program of the American Association for the Advancement of Science grew out of the scientific solidarity between scientists in the 鈥渇ree world鈥� and the 鈥渦nfree world鈥�. Ball has published reports for the AAAS on human rights and databases which are available at under 鈥減ublications鈥�.

How much of your work is dangerous? There鈥檚 a poster on your wall from El Salvador鈥�

Some of it can be very dangerous. That poster was one of a series showing the statistical profiles of the hundred worst officers in the Salvadorean military, with pictures of them and their careers, and the number of human rights violations committed by that unit at that time. That case was dangerous, but it was also fascinating. Since it was during part of the peace process in 1992-93, the officers didn鈥檛 bomb our office or kill us, instead they sued us for defamation. And we鈥檙e saying, like, right on! Let鈥檚 go to court! So we showed up with about 27 crates of testimony, and printouts of all my horribly bad, embarrassing database code. And we said to the judge, 鈥淥K, let鈥檚 go through it.鈥� The judge looked at some of it and called the counsels over and said, 鈥淵ou all should resign.鈥� And they did!

How did you get involved in Milosevic鈥檚 trial?

People at the Human Rights Watch organisation asked me to come to Albania in March 1999 to measure violations suffered by refugees. While I was there, I had several quite heated exchanges with different people from the tribunal 鈥� we disagreed about the relative importance of truth and justice. I argued that it was important that the world, and especially people in Kosovo, should know what really happened during the war and try the people accused of having committed or ordered violations of human rights. But truth should be first. The tribunal鈥檚 field people saw things differently so I did my work independently. Eventually, I went to the tribunal offices, explained what we鈥檇 found and they got interested. They asked what I鈥檇 need to take the case to court. I said: more time, more data and more money. They said that they might be able to help me get more data. They did 鈥� lots of it. We wrote a report linking migration of refugees and killing, using the new data to strengthen the earlier analyses and contradict defence evidence that these deaths were caused by NATO air strikes.

Did you invent human rights data analysis?

Not the statistical models 鈥� they come from other applications, particularly census and wildlife monitoring. But the human rights ideas and some of the particularly complicated pieces of this problem, yes. It turns out that human rights is harder than most other applications because of the complexity of our information streams. When you count gorillas, you don鈥檛 have a lot of different people doing it, and you count them in a tidy set. But in the human rights world, information comes in dribs and drabs in 20 different forms from 20 different sources. Actually, the more they overlap, the denser the information is and the better our analysis tends to be. The problem is, how much have we missed?

What motivated you to go into human rights work?

I became outraged about liberal hypocrisy. It鈥檚 a particular kind of hypocrisy that involves making a set of claims that we are a legitimate state, that America is a good country because we respect these things called rights and we promote freedom. I was particularly outraged by the inconsistency of America鈥檚 actions in Central America in the 1980s.

Where did the computers come in?

There weren鈥檛 that many computers when I grew up in Pittsburgh. My dad was an engineer and worked for IBM in the 1960s, and I grew up in a house full of computer parts. I still use punch cards as notepads. I did a little bit of BASIC programming in high school with a friend of mine who had an Atari. My PhD is in sociology, with a very quantitative bent. I made my living as a statistics and database programmer, writing huge jobs running on IBM mainframes for a social science research house. I put myself through Columbia University doing that in the mid-1980s. When I grokked [understood intuitively] relational databases, I was, like, 鈥淩ight, data鈥檚 not flat.鈥� I took the graduate statistics sequence as an undergrad because I just grooved on it. I thought, 鈥淭his is really cool, we can find stuff out.鈥� So when I realised that relational databases let you look at social space in this multidimensional way, it was amazing.

And at Columbia, you were involved in the protests over the university鈥檚 financial interests in South African corporations?

Yes, I got arrested in divestment demonstrations and into a lot of trouble. And then I was reading Marx, because that鈥檚 what you did if you were a socially progressive lefty kid in the 1980s. A lot of my friends said, 鈥淲ell, if you like revolutions so much why don鈥檛 you go look at one?鈥� So in 1987 I went to Nicaragua for a few months. And I鈥檓 saying, like, 鈥淭his Marxism stuff鈥檚 kind of a bad idea, isn鈥檛 it?鈥� The Sandinistas just didn鈥檛 make sense to me. But at least they were trying to accomplish something good. I wrote them a little database in pirated dBase III.

Grok, groovy, you sound like someone out of the 1960s鈥�

I think those terms come from the geek community. I really like the hippy thing, I feel really attracted to it. Except I have no musical talent and I don鈥檛 really like hippy music. But I like long hair and hanging loose.

But you were born in the wrong decade!

Absolutely. But Hunter S. Thompson was a formative influence in high school. The part I really like is not all the silly drug talk, which I look at as a kind of metaphor for making yourself open, tearing off all your filters so that what you鈥檙e seeing and thinking about just pours into your consciousness. I really liked his book about the 1972 election, which I think is about the end of the 1960s experiment. What I took from that book is, gather every crumb of information that it鈥檚 remotely possible to get your neurons around and then process, then analyse. And that鈥檚 my model for how I do work now.

So how do you go about putting your database together?

Every human rights story goes like this: I am a deponent, and I鈥檓 here to tell you about things that happened to one or many victims. I myself may or may not be one of those victims. Each of those victims may have suffered one or more violations, and those violations may or may not be what historians call colligated at one or more points in time or space. Each of the violations may have been perpetrated by zero, one, or many identifiable perpetrators, and those perpetrators may be individuals with names and ranks, or they may be institutions. Each of those may be associated with one or more of the violations in this story. That鈥檚 the complexity of one story. Now we鈥檙e going to collect 10,000 stories, and there is a dense, complex overlapping of all the stories. Then we aggregate the stories from, say, four different organisations, and each of those organisations鈥� sets of judgements has a dense and complex overlap with the other organisations鈥� information. The result is a multidimensional, multilayered Venn diagram built up from this information, which I refer to as 鈥渞eporting density鈥�.

How do you turn people鈥檚 evidence into a database record?

Our problem is to look at a narrative and extract the countable elements. Enumerate the victims, enumerate the violations. What we do is something called 鈥渁 controlled vocabulary鈥� whereby we build a dictionary of the violations that are interesting to us, and each definition has some basic things it has to do. It has to provide the core content of each violation 鈥� for example, a beating is hitting somebody with a stick or boots. It also has to have a boundary condition, which means that beating is different from burning, or it might or might not include rape. The definition also has to have counting rules: what鈥檚 the difference between one beating and two beatings? I don鈥檛 define how the groups do this, but each group has to agree a definition. The reason you need the detail is because you take 10 copies of the same narrative and give it to 10 people and have them code it and see what the results are: that鈥檚 called inter-rater reliability. It鈥檚 an old problem and there鈥檚 a very old set of statistical methods for a pair of raters, but we generalised it to n raters. So in the Milosevic case, for example, every decision about whether a death was the same in two different databases was made by at least two people and in some cases up to five or six people, and everybody had to talk until they agreed.

You worried about missing information. How much do you miss in a project?

You can model that by looking at reporting density. And the interesting problem is that our coverage across social space is not uniform. We may cover the violations in the cities really well, but out in the countryside we hardly get anything. So what you do is model the reporting density across all the social space and, with some hairy programming, you can see where your lacunae are.

So you have a layered networking approach to data?

Yes: a source layer and a judgement layer. Auditing is crucial. It has to do with the fact that everybody reports overlapping stories. This is a feature, but it looks like a bug to someone who has a simplistic model of what you鈥檙e doing with the database. Does your database faithfully represent your source information or does it faithfully represent your judgement about what is true in the world? If the former, your data will contain both duplicates and lacunae. You adjust by making reasonable estimates of what鈥檚 missing and use other kinds of information that the simplistic model is throwing away.

Throwing away?

Yes! Throwing away information is bad! The other error is to say I鈥檓 going to 鈥渃lean鈥� the data. That phrase makes my skin crawl. It means you鈥檙e going to throw out all the duplicate reporting. Well, that鈥檚 not very respectful to the people who gave you the data, is it? What you need is another database on top of the original database that draws out the elements that we judge to be unique and true. So some elements from the source information don鈥檛 pass to the judgement layer.

Can you spot anything in the careers or biographies of military officers that makes them more likely to violate human rights?

What was the big risk for Salvadorean military officers committing huge human rights violations? US training. It鈥檚 very, very clear. I think US training selects the officers who are the most motivated, and the way that you distinguished yourself in the Salvadorean military in the 1980s was by killing people. So the most motivated officers are also the worst. But what this also says is that US training is useless for restraining human rights abuse. US training also advantages officers relative to other officers, so when they come back they get better jobs and are in a position to commit more violations.

Is free, or at least open-source software important for human rights work?

The other part of the copyright problem is piracy 鈥� or, more appropriately, unauthorised copying. The beauty of this argument, from the point of view of, say, the government of Burma, is that they can say to the World Trade Organization, OK, we鈥檙e going to crack down on piracy 鈥� and then they go arrest a bunch of human rights groups. If we use free software, that all goes away. Also, living in miserable poverty is not a human rights violation, but it鈥檚 clearly not in the spirit of human rights law. I don鈥檛 think any intelligent person thinks that better standards of living are coming to any place without computing. And one of the impediments is paying a huge tax on every computer sold in the world to a rich company in the state of Washington. The way out is a fundamentally different approach to software, but freedom comes at a price. As a human rights guy I accept lower salary, longer hours, more difficult travel schedule and using free software. These are all costs I pay in order to live according to a set of principles.

So what鈥檚 next?

I鈥檓 looking for $15,000 to look at data I鈥檝e had from a long time ago. And right now we鈥檙e taking statements in the truth commission in Sierra Leone. I鈥檓 advising the truth commission about how to deal with all this information, they鈥檙e piling up a ton of statements and we鈥檙e starting to code them.

Do you like this life?

It鈥檚 a big strain on my personal relationships. With all my travel, I have a hell of a hard time getting my yoga practice established as a routine in my life. As a result (of that and a few other unhealthy habits to do with eating animal fats), I have pretty bad repetitive strain injury, I鈥檓 kind of overweight and not very flexible. Going back to Marx, I鈥檇 prefer the kind of life where I鈥檓 an electrician in the morning, a programmer in the afternoon, and a philosopher in the evening. Marx is mostly all wrong, but he had some important insights and one of them is that people want to do stuff that鈥檚 meaningful. I think that work has tremendous dignity, and a good society is one that validates that dignity.