Wondering who will win the next presidential election? Ask Nate Silver.
Wondering which country’s regime could be the next to begin a large-scale slaughter of its own people? Ask Jay Ulfelder.
While data-cruncher Nate Silver’s political prognostications have become the gold standard for American election predictions, political scientist Jay Ulfelder has been working on applying similar statistical techniques to a different, more intense question: Where and when will the next mass killing occur?
In October, Ulfelder and his colleagues will go live with a website that will offer some possible answers to that question. The evolving forecast will be based on carefully interpreted global statistics, as well as the collective wisdom of a crowd of experts who have their eyes on global hot spots.
The website—not yet up at www.earlywarning.org—is the product of the Early Warning Project, an initiative of the U.S. Holocaust Memorial Museum’s Center for the Prevention of Genocide. (Other organizations are engaged in similar efforts, including The Sentinel Project out of Canada.)
Probing the roots of genocide has traditionally been the realm of historians and sociologists. The Early Warning Project, according to Center for the Prevention of Genocide head Cameron Hudson, wants to hand the task over to the people with calculators and spreadsheets. It’s part of a widespread move in recent years toward data-driven narratives, the practice of analyzing huge sets of numbers to answer everything from what a rookie pitcher’s ERA will be to what you should talk about on a first date.
The Early Warning Project started two years ago when Dartmouth professor Ben Valentino, then a fellow at the Center for the Prevention of Genocide, wrote a research paper and held a seminar on the feasibility of a genocide-forecasting model.
Valentino and Ulfelder, both alums of a CIA-funded group working to predict political instability worldwide, got together to put Valentino’s theories into action.
Ulfelder explained the system they’ve come up with, which has two components: a statistics side and a crowd-sourced side. On the statistics side, the system feeds a number of different data points into three separate models. The data includes things like a country’s infant mortality and economic growth rates, whether a government has an authoritarian government, if that government has an exclusionary ideology, and if the country is economically isolated. Based on analysis of historical genocides, all these data points are possible indicators of political instability, or a country ripe for a mass killing event, which for this project is defined as “state-led killing of 1,000 or more non-combatant civilians in sustained violence.”
The three models are then combined, and the system reveals which countries are most at risk of a genocidal event.
The drawback of the statistics side of the project is that much of the needed data is produced only once a year—and a lot can change in a year. “That’s where the opinion pool picks up,” Ulfelder explained.
The Early Warning Project is tapping a crowd of experts—academics, journalists, NGO staffers, regional specialists—and asking them questions along the lines of, how likely is it that, say, Pakistan will see a mass killing event in the next year? The experts can update their answers at any time, responding to changing circumstances on the ground.
The expert opinions are synthesized into another set of predictions, to complement the statistical forecast.
The project already has 115 participating experts. Ulfelder said the goal is to get that number up to several hundred.
Of course, predicting genocide is one thing; preventing it is quite another. The Early Warning Project is designed to create faster and better information so that world leaders and NGOs can take preventative action before a crisis, Ulfelder said. “We’re trying to inform the people who are in a position to make those decisions.”
Preventing or stopping mass killing is clearly a matter not just of adequate information but of cooperation and political will. The Holocaust, Ulfeder pointed out, occurred despite warning signs that were apparent at the time—even without the benefit of modern data analysis.
The project’s goal is to learn from the Holocaust, and the shared traits of other genocidal events, to learn to spot them as they’re approaching, rather than just condemn them after they’ve occurred.
“There’s no reason to think that, while the scale of the Holocaust is certainly unique, that the dynamics leading up to it may be somewhat less unique, and therefore predictable with analysis,” Ulfelder said.