A Warning About Early Warning

Genocide prediction models favored by the Atrocity Prevention Board will not create transparency.

Flickr/Waifer X.The reviews are in and the Obama administration’s Atrocity Prevention Board is getting panned.

When this committee with the mandate to identify future atrocities around the world and recommend administrative responses was rolled out in April, Commentary pronounced it a “silly” and “impotent” vehicle for empty inter-agency posturing. Charles Krauthammer slammed it as bureaucratic farce. James Gibney at Bloomberg View wondered if “the creation of this new entity partly [could] be an old-fashioned bureaucratic power play” and, nodding toward Arendt, called it “the banality of good.” Stephen M. Walt shredded it in Foreign Policy, predicting it will reinforce knee-jerk interventionist arguments without addressing strategic national-interest objectives in a world where American moral policing has narrowed in credibility.

The criticisms, of course, haven’t put an end to the far-flung Atrocity Prevention Board, which is staffed by high-level officials at a dozen federal departments and headed by Samantha Power, the senior director for multilateral affairs and human rights at the National Security Council.

In fact, the board is really only the tip of the spear. In the last several years a much larger effort—galvanized by the U.S. Holocaust Museum, human-rights organizations, academics, and various foundations—has emerged hoping to create early-warning mechanisms to identify and stop genocides and mass atrocities from occurring.

Google.org, the charitable arm of the search engine giant, is funding a project that hunts for surges in hate speech across online social networks. The Genocide Prevention Advisory Network, an informal network of scholars, provides risk analysis and advice to NGOs, the UN, and others. A Canadian non-profit, The Sentinel Project, is building out an extensive atrocity early-warning system. Earlier this month, the UN General Assembly was to convene a special panel, sponsored by the U.S. Holocaust Museum, on the prevention of hate speech and the incitement to genocide (it was canceled due to Hurricane Sandy).

Systems for Prevention

Warning efforts can broadly be divided into two categories. The first type uses quantitative statistical early-warning models beloved by political scientists. These models calibrate risk factors like national poverty and GDP level, regime type, infant mortality (an indication of regime weakness) and experience of armed conflict or political instability. This last variable is really a shortcut since virtually all modern atrocities coincide with an ongoing or a recent history of civil wars, contested elections, coup d’états, or other major political instability. The second type rely on qualitative models that take into account watch lists, findings from human rights groups, media reports and other intelligence.

Countries with regimes now on the radar for future atrocities against their own people include Ethiopia, China, Azerbaijan, Zimbabwe, Colombia, Kenya, Indonesia and Pakistan—none of which, it should be noticed, have any chance of getting invaded by the United States. And early-warning models were prescient in putting Syria near the top of their lists because of its minority-controlled regime and history of internal strife. Burma, which is currently erupting in sectarian violence against its minority Rohingya Muslim population, also set off alarm bells.

But the critiques spurred by the creation of the Atrocity Prevention Board are not focused on the accuracy of a given prediction. In fact, the assumption of accuracy makes their arguments more relevant in assessing this larger wave of early-warning systems. Many of the same objections are in play: What happens when a White House committee or any influential and credible group enshrines the catalyzing phrase “likely genocide” into a policy conversation or a media-hyped national debate? Or when a software program relying on statistical modeling turns guesswork into broad conviction about the unknowable future of human behavior?

Given what will be often interpreted as an imprimatur of certainty, what are the risks of crying wolf? In the hothouse of foreign-policy infighting, will a statistical conclusion on the probability that a government will slaughter its own people survive partisan manipulations? What does any of this matter if it doesn’t change political will?

And perhaps there are already enough experts watching. Don’t we have a massive early-warning system—dub it a “qualitative model” if you want—journalists, human rights investigators and diplomats, as well as bloggers watching satellite images of population movements and weapon transfers? Is early warning even the problem?