Thursday, October 26, 2017

Bayes' Theorem for Intelligence Analysis

Summary and Critique by: Jared Leets

The author, who was an intelligence analyst for the CIA, begins by explaining why intelligence analysts should be interested in probability theory. Intelligence analysis should typically must be undertaken on with insufficient evidence. Bayes' Theorem in its form served participants in the intelligence analyst’s test program as their distinguishing rule for evaluating new evidence and material. In the article the equation R=PL as the odds-likelihood formulation of Bayes' Theorem was used. The author stated that R is the revised estimate of the odds favoring one hypothesis over another, the odds after contemplation of the last piece of evidence. P is the prior estimate of the odds, the odds before contemplation of the last piece of evidence.  Once the estimate was ready, the analysts participating did not make any judgments regarding P. The participating analysts gave only insight about about L, which happened to be the likelihood ratio. The author stated that the likelihood ratio was the analyst’s assessment of the distinguishing piece of evidence.

The author stated that there exists three features of Bayesian probability that differentiate it from conventional intelligence analysis. The first is that the intelligence analyst is asked to quantify judgments which he or she does not typically do in numerical terms. This feature is what draws the most criticism against Bayesian in the intelligence community. Analysts must disagree in their opinions of the exact figure that shows the distinguishing value of a piece of evidence. A proponent of Bayesian might say that disagreement among analysts is simply a characteristic of traditional method and is no less serious for being implicit compared to being explicit in the intelligence analysis. Usually it will be challenging for most intelligence analysts to use, as most are not that mathematically capable and cannot express degree of belief to the precision inferred by the numerical value.

Another feature discussed in the article was the feature of Bayesian method that the analyst does not take the available evidence at face value and then draw conclusions regarding the merits of opposing hypotheses. Zlotnick states, “He rather postulates, by turns, the truth of each hypothesis, addressing himself only to the likelihood that each item of evidence would appear, first under the assumption that one hypothesis is true and then under the assumption that another hypothesis is true. He does not feel called upon to reinforce his self-esteem by reaffirmation of opinions previously put on the record.”

Finally the last feature of Bayesian method is that the analyst comes to a conclusion on evidence given. The analyst does not add the evidence as he or she would typically do to judge its meaning for the final product. The math does the summing up, saying to the analyst that these pieces of evidence equal this conclusion. Research suggests that analysts are better at distinguishing a single item of evidence than at drawing inferences from the evidence as a whole.

Bayesian could possibly complement intelligence analysis when using it for strategic warning analysis due to the fact that it resolves around the odds favoring one hypothesis (say imminent attack) over another hypothesis (no imminent attack). Other ways to test Bayesian in the intelligence analysis field is to test it on international crises from the past. This is done by reviewing evidence from the past and looking at how they made estimates based on their evidence. In the end there must be an assignment of L values and likelihood ratios.

Overall this was article an excellent article explaining how intelligence analysts were attempting to incorporate Bayesian into their analysis. Probability theory is quite relevant in the intelligence community. The author did a good job of explaining what Bayesian is and how it could be used for intelligence analysis. While the author did discuss the strengths of using it, he also spoke of its weaknesses. He looked at several ways that it could be used.



  1. Bayesian inference is a subjective and alternative way of looking at it, which may be taken into consideration when added information is available, in order to mathematically quantify the analyst’s perceived risk. As the author said in his research, “Bayesian method is a mathematical logic to which intelligence can have recourse for substantiating or contradicting the verbalizations of the traditional analysis”. In other words, it gives the analyst the ability to quantify the data, analyze it, and give an estimated assessment for their superior. In the end, I agree with the author that the world is full of fallible judgments about evidence and that the Bayesian approach is not a path to perfection but it can be a path to make it clearer.

  2. The biggest flaw within the methodology in relation to intelligence analysis is the major emphasis on numbers. It does not take into account individuals with minimal math knowledge.

  3. In intelligence analysis, data is often unstructured, incomplete, or deceptive. This often makes traditional statistics an unrealistic method. Nonetheless, it remains more important than ever for analysts to be able to quantify, update, and defend their judgements given the unforgiving nature of intelligence failure. Although Bayesian statistics is often disregarded as too mathematically complex, it would be interesting if the intelligence community could formulate a more simplified methodology for its analysts.

  4. To use Bayes effectively, the subjective nature must be tempered. While it is a flexible methodology in that it allows for new evidence, the amount of weight given to new evidence can be subject to bias based on prior knowledge.