Intelligence_ From Secrets to Policy - Mark M. Lowenthal [97]
If a lesson is to be drawn from these two analytical experiences it may be no more than that the analytical process is imperfect under any and all conditions. No Goldilocks formula has been devised as to the right amount of intelligence on which analysis should be based. The quality of that intelligence matters a great deal, as does the nature of the issue being analyzed.
CONVEYING UNCERTAINTY. Just as everything may not be known, so, too, the likely outcome may not be clear. Conveying uncertainty can be difficult. Analysts shy away from the simple but stark “We don’t know.” After all, they are being paid, in part, for making some intellectual leaps beyond what they do know. Too often, analysts rely on weasel words to convey uncertainty: “on the one hand,” “on the other hand,” “maybe,” “perhaps,” and so on. (President Harry S. Truman was famous for saying he wanted to meet a one-handed economist so that he would not have to hear “on the one hand, on the other hand” economic forecasting.) These words may convey analytical pusillanimity, not uncertainty. (Conveying uncertainty seems to be a particular problem in English, which is a Germanic language and makes less use of the subjunctive than do the Romance languages.)
Some years ago a senior analytical manager crafted a system for suggesting potential outcomes by using both words and numbers—that is, a 1-in-10 chance, a 7-in-10 chance. Such numerical formulations may be more satisfying than words, but they run the risk of conveying to the policy client a degree of precision that does not exist. What is the difference between a 6-in-10 chance and a 7-in-10 chance, beyond greater conviction? It is also important to remember that an event that has a 6-in-10 chance of occurring also has a 4-in-10 chance of not occurring. When presented this way, the event now may seem uncomfortably close to 50/50, which a 6-in-10 chance does not convey by itself. There are very few “sure things.” In reality, the analyst is back to relying on gut feeling. (One chairman of the NIC became incensed when he read an analysis that assessed “a small but significant chance” of something happening.)
One way to help convey uncertainty is to identify in the analysis the issues about which there is uncertainty or the intelligence that is essentially missing but that would, in the analyst’s view, either resolve the unknowns or cause the analyst to reexamine currently held views. This raises another issue: known unknowns (that is, the things one knows that one does not know) versus the unknown unknowns (that is, the things one did not know that one did not know). By definition, the second group cannot be bounded or reduced as it is unknown. But one’s analysis must constantly be examined to identify known unknowns and to give attention to resolving these issues, if possible.
The use of language is important in all analysis. Analysts tend to use a stock set of verbs to convey their views: “believe,” “assess,” “judge.” For some analysts the words have distinct and separate meanings that convey the amount of intelligence supporting a particular view and their certainty about this view. However, the intelligence community did not reach a consensus as to what each verb meant until 2005. The NIC now publishes an explanatory page with each NIE that explains the use of estimative language. The text box, “What We Mean When We Say: An Explanation