You might be surprised to hear “there’s an app for that” for an issue as heavy as genocide prevention. But the Simon-Skjokt Center for the Prevention of Genocide, overseen by the US Holocaust Memorial Museum, has released just that—a beta of the “Early Warning Project,” described by director Cameron Hudson as an “analytic framework for both what is driving risk and the nature of risk” of genocide in all the countries of the world.
Since the creation of the Holocaust Museum 21 years ago, it has been dedicated to remembrance and learning from the past. According to Hudson, “until now any kind of forecast has been prognosticating. It’s been qualitative—people saying this or that might happen, and while [the project] is not meant to replace the tried-and-true methods of prognostication, it’s another data point for people to think about.”
“The hope of this project isn’t so much that people focus on the top 10 list of countries at risk; it’s countries in the 20-30 range where there’s no policy focus and no attention being paid—such as Sri Lanka or Bangladesh or Guinea or Tanzania,” Hudson says. “This year such a country might be [number] 30 on the list, but next year it could be number five.”
A similar program is already used within a CIA project called the Political Instability Task Force. Data Scientist Ben Valentino of Dartmouth sat on their board as an academic advisor before transitioning into the Early Warning Project as one of their project fellows. The project started two years ago out of a desire to create publicly available predictions from publicly available data from agencies like the World Bank and United Nations.
“Right now, 99% of what our government and other governments do is crisis response,” Hudson says. “By the time shots are fired and body bags are being lined up, you’re no longer in preventative mode. [This program] is a way to put a canary in the coal mine where we don’t currently have one.”
Dozens of trends ranging from GDP and income inequality to regime type and ethnic diversity in society and government have been tracked annually since the end of World War II. Using a couple of statistical models, these factors all come to annually predict the likelihood of genocide. The software also employs machine learning to develop new statistical models by recognizing trends which until now have not been realized to be important.
Because these datasets are updated only once a year, the Early Warning Project also employs a separate expert crowdsourcing program to capture emerging threats that occur throughout the year. Thousands of people anonymously fill out surveys concerning perceived risk of genocide.
“We often see an uptick after a coup—we saw that in Burkina Faso recently,” Hudson explains. “A few months ago the government of Burma kicked out international aid organizations that were providing assistance to a Muslim minority called the Rohingya who have been targeted for persecution.”
Perceived risk went from 15 percent to over 50 percent. “I think if you’re a policy maker, that’s a useful data point to have,” Hudson says. “If there’s a group of people paying attention to these kinds of things that care about these issues whose concern has gone up significantly, that might be something you want to dig into.”
There are three groups of people the Early Warning Project makes an effort to incorporate into expert crowdsourcing: people with regional expertise, people with functional expertise in genocide prevention, and people who are close to the conflict zones. Having a substantial fraction of each has been insightful. “We ran a question recently on Zimbabwe,” Hudson says. “People in Africa perceived the increase in risk to be significantly lower than people who were responding from Europe or the United States. We’re not sure what to make of that other than that it’s an interesting disparity.”
But Hudson is hopeful. “There’s a lot more that we don’t know for certain than what we do know for certain. Hopefully this tool, as it runs, will help shed light on those things that we just don’t know for certain.”