Today on the AIPR Blog, Daniel Solomon, an independent researcher on mass atrocity issues who blogs at Securing Rights, discusses The Early Warning Project, a new mass atrocity forecasting program that combines statistical forecasting with crowd-sourced intelligence from a pool of invited experts, of which Solomon is among.
In December 2008, the final report of the Genocide Prevention Task Force (GPTF), a high-profile convening of U.S. officials and policy experts, described “early warning” as a prerequisite of effective mass atrocity response. Early warning, which the group defined too-narrowly as “getting critical information [about mass atrocities] to policymakers,” has since emerged as a keystone of local, national, and international mass atrocity prevention agendas. The first adage of mass atrocity prevention, that it is possible, is closely followed by a second: mass atrocities can be known, far in advance of their onset.
Early warning, defined more broadly, far predates the GPTF report; since 2008, however, policy-oriented warning programs have proliferated widely. Recent innovations in statistical forecasting have borne fruit in parallel models of mass atrocity risk. Among several programs, one stands out: the Early Warning Project, a partnership between the U.S. Holocaust Memorial Museum’s Center for the Prevention of Genocide and Dartmouth’s Dickey Center for International Understanding. Unlike similar programs, which rest on the predictive grist of specific models, the Early Warning Project creates an interaction between statistical forecasting and the “wisdom of the [expert] crowd.” In the project’s expert opinion pool, of which I am a member, area scholars and mass atrocity specialists fill the blind spots of quantitative indicators. At the same time, the project’s quantitative side informs the specialists. If we can anticipate mass atrocities before they occur, it is only through the pluralism of knowledge, which the Early Warning Project advances.
Like much of the mass atrocity prevention agenda, early warning does not exist in a vacuum; each program warns about a specific event, for a specific audience. Some programs, formal and informal, identify local threats to vulnerable civilians: in northern Nigeria, an interfaith consortium gathers town-level data about trends in mass violence, for town-level use. The scope of the Early Warning Project is general by design. The project’s statistical model—and, to a partial extent, its expert opinion pool—measure risk in “country-years,” namely, annual, country-level indicators of mass violence. In tandem, the model and the expert pool can describe the risk of mass violence in Sudan, in 2015; only the pool, with its malleable question set, can describe the same risk in Darfur during November of the same year. The predictive model is only as nuanced as its data allows. Without more granular indicators, more specific prediction exceeds the project’s scope. It leaves to others—field researchers, journalists, human rights monitors—the question of how perpetrators may kill, or where, or when.
If the project’s audience is diverse, it is also limited. The explicit audience of the project’s warnings is a digital public: an assortment of individuals, organizations, and officials, often linked to institutions of the so-called global North, with continuous access to web-based information. This is a significant improvement from the status quo. Existing attempts at systematic warning models, such as those spearheaded by the CIA-funded Political Instability Task Force, are proprietary to the U.S. government; separate public efforts to assess and rank the global risk of mass violence, such as the International Crisis Group’s CrisisWatch bulletin, often lack comprehensiveness, whatever their qualitative value. Despite these improvements, the shortcomings of the program’s audience are also clear. These warnings are not intended for communities in conflict-affected areas. In fact, civilians and civil society for whom violence is imminent will likely find little use for the program’s country-year assessments. These groups may receive separate, more specific warnings that describe the time and location of a society’s mass violence.
In some circumstances, a common practice of mass atrocity prevention is unachievable and unproductive. As AIPR Board Member Sheri Rosenberg often observes, the relative entropy of the field of mass atrocity prevention may prove more fruitful than its false organization. A practitioner may approach the prevention of mass violence from various angles and levels of political organization, each equally as worthy as its counterpart.
Several approaches to prevention, however, may benefit from the agenda the Early Warning Project’s risk assessments provide. Where it works best, the Early Warning Project is an agenda-setting tool: it tells practitioners, if imperfectly, which crises loom on the near horizon. For the informed practitioner, the program’s findings will contain few surprises (any way you slice it, the likelihood of mass violence in countries like Myanmar is an apparent fact). Its strength is not in the creation of new knowledge, but in the transparent collection of existing knowledge about global trends in mass violence. The statistical model’s criteria is publicly accessible: through its blog, the project is also finding ways to make the conclusions of its expert pool equally transparent. The advocacy efforts of coalition groups like the Prevention and Protection Working Group, which informed the creation of the U.S. government’s Atrocities Prevention Board, or the R2P Focal Points initiative of the Global Centre for the Responsibility to Protect, which organizes global collective support for the “responsibility to protect” doctrine, are often ad-hoc. These groups may use the Early Warning Project’s findings, as well as associated media coverage, as a wedge for collective global and regional action.
Of course, the political tenuousness of prevention will remain. One imagines the government of South Sudan, for example, is none too pleased by its unfortunate rank on the Early Warning Project’s list. But the possibility of limited consensus, often elusive in the practice of mass atrocity prevention, will likely advance our current status quo.