The World

Can You Create a Formula to Predict Mass Murder?

Guard towers and barbed wire at Auschwitz on Jan. 25, 2005.

Photo by Sean Gallup/Getty Images

“Genocide can destroy a culture instantly, like fire can destroy a building in an hour,” wrote Raphael Lemkin, the Polish-Jewish attorney who coined the term genocide and led the campaign to have it banned under international law after World War II.

While mass atrocities and genocides can occur quickly, they don’t come out of nowhere, with warning signs often evident years in advance. How predictable are these warning signs? Is it possible to discern where the next Nazis, Khmer Rouge, and Hutu Power will emerge?

A new initiative from the U.S. Holocaust Memorial Museum is looking to answer that question. The Early Warning Project, a collaboration between the D.C.-based museum and the Dickey Center for International Understanding at Dartmouth, is now rating countries on the likelihood of state-orchestrated mass atrocities. “The U.S. government has had estimates like this for a while, but they haven’t made them public. So part of this is to democratize this information and put it in the hands of public advocates, journalists, and others,” says Cameron Hudson, director of the Holocaust Museum’s Center for the Prevention of Genocide.

The organizers say the full site will be operational in a few weeks. For now, you can check out the project, including country ratings for 2014, on a WordPress blog.

The ratings combine assessments from a number of statistical models, rating political instability, the likelihood of civil wars and coups—mass atrocities are often preceded by violent changes of government—as well as economic factors and the ethnic composition of a country’s leaders. The site will combine these rankings with aggregated polls of experts to highlight countries where mass killings may occur. A recent expert roundup, for instance, examined the likelihood of a mass killing in Russia’s restive North Caucasus region.

The 2014 list is led by a few predictable countries. The gray dots show the scores from the different statistical models that are being combined. The red dot is the average. The X axis shows the probability of an event taking place, so 0.15 equals 15 percent:

Myanmar (aka Burma), Sudan, and the Central African Republic are all places where atrocities against civilians have taken place recently. It’s not a stretch to think such events could happen again. Jay Ulfelder, a political scientist who helped develop the model, says this is a feature, not a bug. “Most of the time, the countries at the top of the list are not going to be surprising,” he says. “That’s good. If it were totally counterintuitive, we’d probably have a bad model.” Ulfelder, who’s also known for developing a model for predicting the likelihood of military coups, says “the value comes from the small number of cases that are counterintuitive.” Guinea, for instance, ranked just above Afghanistan, is “one that people in the atrocity prevention community aren’t talking about.”

Guinea’s inclusion makes some sense when you look at the underlying explanations. The country experienced a military coup in 2008 and has some ongoing ethnic violence. (The 2014 Ebola outbreak does not factor into Guinea’s ranking, as most of the data the Early Warning Project are using at this point comes from 2012 and 2013.)

Rwanda’s high ranking is another surprise. Though the country obviously has a history of mass atrocity, it’s been relatively stable and prosperous under Paul Kagame’s increasingly authoritarian but internationally lauded government. Nonetheless, the ethnic divides that led to the country’s genocide are still present, and fractures in the ruling elite or discontent from the population could reignite them. As Ulfelder puts it, “Rwanda has taped over those issues but hasn’t really resolved them.”

It’s also interesting to see democratic India near the top of the list given recently elected Prime Minister Narendra Modi’s controversial role in Gujarat’s religious riots in 2002. Modi, once a hard-line Hindu nationalist, was chief minister of the state during riots that caused the deaths of hundreds, mainly Muslims. He is blamed by some for not doing enough to stop the violence, or even encouraging it, though he has denied these charges and has been cleared of any wrongdoing in court. Ulfelder says the score isn’t driven by Modi but has more to do with the conflict over Kashmir, where “the historical model we’re using sees state-led mass killing until fairly recently.”

On the flip side, Zimbabwe, which saw mass killings during the disputed election of 2008 and where tension and uncertainty surround the question of who will succeed 90-year-old dictator Robert Mugabe, the world’s oldest president, gets a surprisingly low rating using the model. According to Ulfelder’s analysis, Zimbabwe’s simmering political tensions could push it higher in a future assessment, but for now, it doesn’t have the overt crises that have pushed up the scores of other countries. The risk is there but it is not as imminent.

The model only measures the likelihood of mass atrocities perpetrated by governments. Iraq is high on the list not because of the massacres of civilians, particularly ethnic minorities, perpetrated by ISIS this year. Rather, it’s because of the high likelihood of atrocities by Iraqi government forces and associated Shiite militias—atrocities that are likely already taking place.

Hudson says part of the goal of the project is educational, in keeping with the Holocaust Museum’s mission. “Education isn’t just about the Holocaust, it’s about what causes genocides to begin with,” he says. “What happened in the years from the 1920s to 1942 is much more useful for historical analysis than just the three years of mass killing. It’s a much more powerful story to tell in terms of how these events happen and how to prevent them.”

He also hopes the site will help human rights groups flag potential trouble spots before killings begin. “What we’re trying to do is move the conversation away from genocide response, which is all we talk about in Iraq, Syria, and Sudan, you name it, to start to have thoughtful conversations about a whole host of countries that are not in conflict but could begin to experience this kind of crisis,” he says. Hudson adds, “Avoiding a heart attack is better than going in for open heart surgery.”

The creators don’t claim that the model has anywhere near pinpoint accuracy. It can’t tell you when a potential Hitler or Pol Pot will translate words into actions. But it is a way of more empirically assessing an atmosphere of potential violence. This isn’t the first effort at data-driven atrocity prediction—a project from the Canada-based Sentinel Project that I wrote about last year, for instance, employs user-provided data to map the amount of hate speech in a country’s media. The advantage of the Early Warning Project is that it combines a number of factors. To return to Hudson’s medical metaphor, it can’t tell you that the patient is about to have a heart attack, but it can give you a pretty accurate measure of his cholesterol, weight, and stress level.

Any discussion of the international response to genocide is going to raise the controversial issue of humanitarian military intervention. The Obama administration, which has generally favored limited strikes against terrorists via drones or special forces over large-scale military interventions, has made exceptions in cases where genocide appeared likely. Obama justified airstrikes against Libya in 2011 on the grounds that they prevented an imminent “massacre” in Benghazi. The plight of Yazidi civilians surrounded in northern Iraq helped prompt the ongoing military operation against ISIS that began last summer.

The U.S. isn’t the only country making these judgment calls. France, for instance, has in recent years deployed peacekeeping troops to bloody conflicts in its former colonies the Ivory Coast, Mali, and the Central African Republic.

There is no single algorithm that can tell governments when and how to step in to prevent bloodshed. But the Early Warning Project’s designers do see it as part of a toolkit to help make those calls. “We’re not in the business of making explicit policy recommendations, but we hope to inform those decisions,” Ulfelder says. “On balance, having better information available should make those decisions better.”