Also, even when people feel they know nothing, they typically know a bit and that bit should tip them away from maximum uncertainty, at least a bit. The astrophysicist J. Richard Gott shows us what forecasters should do when all they know is how long something-a civil war or a recession or an epidemic-has thus far lasted. The right thing is to adopt an attitude of "Copernican humility" and assume there is nothing special about the point in time at which you happen to be observing the phenomenon. For instance, if the Syrian civil war has been going on for two years when IARPA poses a question about it, assume it is equally likely you are close to the beginning-say, we are only 5% into the war-or close to the end-say, the war is 95% complete. Now you can construct a crude 95% confidence band of possibilities: the war might last as little as 1/39 of 2 years {or less than another month}, or as long as about 39 × 2 years, or 78 years. This may not seem to be a great achievement but it beats saying "zero to infinity." And if 78 years strikes you as ridiculously long that is because you cheated by violating the ground rule of you must know "nothing." You just introduced outside-view base-rate knowledge about wars in general {e.g., you know that very few wars have ever lasted that long}. You are now on the long road to becoming a better forecaster. See Richard Gott, "Implications of the Copernican Principle for Our Future Prospects," Nature
( Philip Tetlock )
[ Superforecasting: The Art and ]
www.QuoteSweet.com