How to Think When the Evidence Is Incomplete

The inquiries into the intelligence failures preceding the 2003 invasion of Iraq left me with a question that goes beyond that particular case: how do we prevent our assumptions from quietly becoming our conclusions?

A useful book to approach this problem is Randolph H. Pherson and Richards J. Heuer Jr.’s Structured Analytic Techniques for Intelligence Analysis (third edition, CQ Press/SAGE, 2021). Its starting point is familiar to anyone who works with intelligence: information is often incomplete, ambiguous and sometimes deliberately misleading. Analysts still have to reach judgments, frequently under time pressure. The quality of those judgments depends in part on whether their reasoning can be examined and challenged.

The authors present 66 structured analytic techniques designed to make the steps of analysis more explicit. The Key Assumptions Check asks which untested beliefs support a conclusion. Analysis of Competing Hypotheses requires analysts to consider alternative explanations and pay particular attention to evidence that contradicts them. Premortem Analysisasks a team to imagine that its assessment has proved seriously wrong and then investigate why. These methods encourage analysts to expose disagreements and test the logic connecting evidence to conclusions.

I find the emphasis on disconfirming evidence particularly valuable. It is easy to collect facts consistent with an initial hypothesis; the more difficult task is to ask what would make that hypothesis untenable. An item of information may appear persuasive until we realise that it is equally consistent with several different explanations.

The book explicitly connects the development of these techniques to the intelligence failures surrounding the attacks of 11 September 2001 and the erroneous 2002 assessment of Iraq’s weapons programmes. Yet its authors are careful about the limits of method: no technique guarantees the right answer, and better reasoning cannot manufacture missing evidence. Expertise, imagination and an institutional environment open to criticism remain essential.

I would recommend the book to students well beyond intelligence studies. The same questions arise in legal research, public decision-making and, increasingly, when using artificial intelligence: What do we actually know? Which assumptions are doing the work? What evidence would change our minds? A convincing conclusion is worth more when we can see how it was reached.

Reading: Randolph H. Pherson and Richards J. Heuer Jr., Structured Analytic Techniques for Intelligence Analysis, 3rd ed., CQ Press/SAGE, 2021. Particularly useful for students: Chapters 2 (The Role of Structured Techniques) and 7 (Diagnostic Techniques).