Saturday, February 9, 2008

Human Rights with Statistics

A human rights statistician finds truth in numbers
Mr. Ball was the first expert witness called in the case against the former Serbian president, who was representing himself against mass atrocity charges at the International Criminal Tribunal for Yugoslavia. Ball had spent 10 months crunching numbers about migration patterns in the former Yugoslav province of Kosovo; his findings suggested that hundreds of thousands of refugees who fled to Albania were spurred by the violence of Mr. Milosevic's army. By the time Ball entered the tribunal chamber, in March 2002, the ousted leader had a reputation for grand orations rather than direct questions; when Milosevic veered off track, the judge would interrupt. "Milosevic would say, 'Dobro,' and go on...." Ball remembers. "It means, 'OK, very well,' but it was clearly a, 'Very well, we'll have you shot later.' I hear [that] in my dreams periodically."

Ball is a statistician – not exactly a profession usually associated with human rights defense. But the Human Rights Data Program that he heads at Benetech, a technology company with a social justice focus, is bringing the power of quantitative analysis to a field otherwise full of anecdote....

"The thing about human rights violations is that they occur massively. They don't occur one at a time," he says. What turns out to be really important, he says, is whether it's thousands or tens of thousands. "Because ... we have very different political understandings of [numbers]."

Since 1988, Ball has been "hacking code" – writing software – to unlock secrets from numbers. He taught himself computer programming so he could get a job that would cover expenses not included in his undergraduate scholarship to Columbia University. Not much of a campus radical, he did earn four years of disciplinary probation for helping to chain shut the doors of a building, hoping to pressure the university to divest holdings in companies doing business in then-apartheid South Africa.

He wouldn't find himself on the front lines of human rights work until grad school at the University of Michigan in the late '80s, when the Central America crises were hot campus topics. All the talk felt empty to him: "When you're in a university in North America ... you're learning about all this stuff you can't do anything about.... You can have these stupid little campus demonstrations, but who are you talking to?"

He took a leave of absence and went to El Salvador with the Peace Brigades, an international group that offered foreign escorts to high-profile local leaders. He liked the idea that guerrilla fighters or government soldiers might be less inclined to commit atrocities in front of Western witnesses. But as the war wound down, he felt less useful. When a human rights commission asked him to do some computer work for them, he was relieved. "Accompaniment was boring," he says, "and programming was fun."

Ball wrote software that allowed the commission to aggregate and analyze the human rights records of officers in the El Salvadoran Army. The results forced a quarter of the military leadership to retire....

Like all statisticians, Ball began with the most basic hypothesis: In looking for a common cause, he is already wrong. Statistics begins with an original assumption – that everything is random – and discards it only when the data suggest otherwise. In Ball's case, they did: He found patterns in the mass movement of refugees strong enough to suggest that more than ordinary wartime chaos was at work. At the same time, the relationship between migration and NATO or KLA actions was so weak that he knew neither was the cause.

Statisticians have a language for description without interpretation. When the analysis showed the movements were neither random nor likely to follow NATO or KLA activities, Ball wrote: "The migration patterns of Kosovar Albanians are consistent with the hypothesis that there was a coordinated and organized effort to drive them from their homes." In layman's terms, the data suggested ethnic cleansing. In fact, the migration patterns matched killing patterns "so unbelievably perfectly" that he concluded that the two situations might be explained by the same external influence.

But this is where statistics, a science of elimination, cedes to lawyers, human rights practitioners, and historians. Observing a "consistent hypothesis" isn't the same as naming a cause. "When we're looking at data, it's what we're able to observe. That's not the same as what is true."

In the end, Ball can't say what did happen; he can only estimate what probably didn't. But even this reveals something bigger about the nature of truth: At a micro level, it seems to change, from town to town or person to person.

In Peru, Ball's team estimated the dead or disappeared in that nation's terrorist war in the 1980s to be twice as high as the estimate made by a human rights commission in Lima. "They said, 'How did we get it so wrong?' " he recalls. "Your risk of being killed ... up in the Andean highlands was 400 times greater than your risk of being killed in Lima ... [where] you feel like you're in the war, but ... a completely different war than the people up in the mountains."

Patrick Ball interview
Route 66

No comments: