Sadly, the conclusion that Green and Armstrong reach-that experts should not be used for predicting conflict outcomes-is not a surprise. Decades ago, Armstrong taught us that expertise beyond a minimal level does not improve judgmental accuracy across a variety of domains (Armstrong 1980). More recently, Tetlock (2005) drove home that point in a study of hundreds of political experts who made thousands of forecasts over many years. Like Green and Armstrong, Tetlock found the expert forecasts to be frequently inaccurate. In support of Armstrong's previous work, Tetlock suggests that avid readers of The New York Times should be able to predict political events as well as highly trained experts. Green and Armstrong also demonstrate that non-professionals mistakenly expect superior performance from experts relative to what they expect from novices. Although it is true that neither novices nor experts were more accurate than chance in eight conflict-prediction tasks, most study participants did not begin with high expectations of the experts in the first place. Participants expected experts to be accurate 45 percent of the time in tasks in which random guessing would yield a success rate of approximately 28 percent. Although these expectations were higher than chance, they are hardly a ringing endorsement for the perceived value of expert forecasters.
展开▼
