In the picture
Cover of Jeffrey A. Friedman's book 'War and Chance. Assessing Uncertainty in International Politics' (New York: Oxford University Press, 2019), 228 pages.
Decision-making by a president or a government requires clearing up uncertainties about the risks involved in such actions. Intelligence services have a long tradition of estimating the probability that a piece of information received is true or that what analysts think might happen will happen. A scene from the movie 'Zero-Dark-Thirty', about the U.S. Navy Seal operation against Obama bin Laden, reflects a typical status : around a table, the different security agencies tell President Barack Obama the percentage by which each of them estimates the probability that the Al Qaeda leader actually resides in the house in Abbottabad that has been placed under surveillance. Obama later said he considered the wide range of estimates given to him by the various agencies to be unhelpful because of their dispersion; nevertheless, he ended up ordering the attack based on the assurances of success given to him by the military, also in an attempt to remove uncertainties, in the event that Bin Laden was indeed in that Pakistani hideout.
That episode appears in several passages of 'War and Chance. Assessing Uncertainty in International Politics'. The purpose of the book is to raise the convenience that the "probabilistic reasoning", frequent in intelligence analysis, transcends that reduced field and is also applied to decision making in the broader area of the International Office. Jeffrey A. Friedman, researcher specializing in the art of prediction ('forcasting') and promoter of several experiments carried out with Philip Tetlock (author of 'Superforcasting'), regrets that international policy analysts most often remain at the stage of merely describing the forces at work or that, even when attempting to peer into the future, they limit themselves to the presentation of various possible scenarios, usually focusing on the worst-case scenario, rather than determining the probabilities of each scenario occurring. At other times, analysts advocate the implementation of policies based not on the reality that is most likely to occur, but on the scenario that would be desirable.
War and Chance', which despite degree scroll is not really about war (this is taken as the epitome of foreign relations), is a highly technical work, aimed at an audience accustomed to quantitative analysis in the social sciences. Before this audience, the author tries to justify the application of probabilistic reasoning even to questions in which there is no room for scientific measurement.
Friedman is at agreement with Isaiah Berlin's statement that "to ask for or preach mechanical precision, even in principle, in a field incapable of it is to be blind and to deceive others". So what he advocates is the perception of "subjective probability": by assessing the ascertained facts and examining the factors that determine possible developments of events the expert analyst may be able to estimate the probability of their occurrence. Even if the reasoning is based on statistical analysis or mathematical models, in the end the analyst's judgment will reflect personal convictions rather than objective truths.
In his experiments, Friedman found that "when a wide range of analysts took the time and effort to assess uncertainty in a clear and structural way, then the accuracy of their judgments consistently improved". Not that the policymaker or any other qualified decision-maker should receive that precise evaluation from the analyst as a certain oracle, but it would be absurd not to take it into account, among other elements, when it comes to limiting the surprise of the unknown. "Assessing uncertainty is not enough to make reliable foreign policy decisions, but it is a necessary component for it," says the author.
Friedman counters various criticisms of this commitment to probabilistic reasoning, among them the warning that determining the probability of an event using a scale that goes from 'remote' to 'almost certain', passing through 'improbable' or 'highly probable' -especially if its translation into probability percentages is applied- can give the "illusion of scientific rigor" when in reality it is a subjective assessment, however well founded it may be. Friedman replies that to the extent that analyst and decision-maker are aware of the limitations of knowledge in which they operate and of the significance of the levels of the probabilistic gradation used, such precision when communicating foresight is beneficial. This is much more useful, according to the author, than the vagueness that many analysts incur when looking ahead, leaving the addressee of their analysis without guide in their decisions.
Friedman summarizes in four points what would constitute the basic standard of the foresight analyst's work for assessing uncertainty:
1) Foreign policy analysts should describe the uncertainty surrounding any predictions or policy recommendations they make. It is especially important to describe the likelihood that the policies being recommended will achieve their intended objectives. This should be seen as required by an analyst's own rigor.
2) Analysts should never base their predictions or policy recommendations using "relative probability" or "conditioning" (the Pentagon based its recommendation to Obama for a troop increase in Afghanistan on the fact that the status would clearly improve over what it was at the time, but without quantifying what the chances of ultimate success were).
3) Estimates should be sufficiently clear for the receiver to understand the precision to be conveyed. This is best done by providing numerical percentages; however, if an analyst prefers qualitative language, at least ten probability levels should be used to ensure accuracy.
4) Distinguish clearly between probability and confidence estimates: beyond indicating how likely a given development is, it is useful to describe how this conclusion was reached, showing how much confidence is placed in this prospect.
Although the author presents this as a standard, in reality it should be received as a maximum approach, a horizon to which it would be convenient to tend. In the field of intelligence there is a long internship of probabilistic reasoning, perhaps facilitated by the need to estimate the validity of each of the information obtained and handled. In other fields of analysis it is not so simple; it is not, for example, to estimate the Degree of success that a certain decision can reach, because it is less possible to reduce the validity of the eventual scenarios to probabilities of occurrence. Nevertheless, it is advisable for analysts to include more foresight in their analyses and to do so with greater rigor.