The Early Warning Project does not try to assess risks of genocide. Instead, we assess risks of mass killing—a closely related but broader concept that covers all of the violent episodes that most observers would consider to be genocides, along with many others in which large numbers of civilians were deliberately killed.
Our decision to focus on mass killing instead of genocide is, in part, a practical one. We use statistical models in our risk-assessment process, and these models must be trained on historical data. If the number of events of concern in our historical data is very small, those models will have a hard time discerning useful patterns. Training models on a data set with very few examples is a bit like trying to make out images in a pointillist painting with only a few dots. We need more information to see the contrasts that interest us.
As it happens, the set of historical cases that most observers would consider to be genocides is quite small. In her foundational work on the topic, Barbara Harff broadened the concept of genocide to include efforts to destroy political as well as national, ethnic, racial, or religious groups and still identified only 36 instances worldwide between 1955 and 1997. The fact that examples aren't more abundant is a good thing for humanity, of course, but it does make statistical forecasting of genocide exceptionally difficult.
Bosnian Muslim women at a Sarajevo protest on June 11, 2007 (via the Srebrenica Genocide Blog)
Politics does play a role in our decision, too, though. In a recent brief for the Carnegie Endowment for International Peace, Matthew Kupfer and Thomas de Waal remind us that,
The word genocide has been used as a weapon of political rhetoric for more than sixty years. Since its coinage in the 1940s, in popular political vocabulary—if not in international legal circles—the term genocide has been used as a signifier for "ultimate evil.”
The intense politicization of the term "genocide" plays into efforts to prevent or mitigate its occurrences in complicated ways. In a December 2013 blog post, Louisa Lombard described how this politicization had factored thus far into international responses to the conflict in the Central African Republic (CAR).
Kupfer and de Waal see similar dynamics in rhetoric around the ongoing war in Ukraine:
As these examples suggest and Lombard surmises, this labeling exercise and the arguments that inevitably ensue don't always advance the cause of preventing atrocities.
The Early Warning Project is not an advocacy group. Our expertise is technical. We aspire to inform policy and advocacy that may help to prevent atrocities by producing risk assessments that are as reliable and timely as possible.
We believe we can fulfill that warning mission more effectively if we do not entangle ourselves in acrimonious debates over which cases deserve to be described as examples of the "ultimate evil" and which do not. Our risk assessments and observations are inherently political, but calling something a "mass killing" does not invoke the same intensity of feeling or have the same international legal implications as calling it a "genocide" would. The pragmatic demands of statistical modeling already push us away from focusing on genocide, but these political sensitivities only reinforce that decision.