Analysis of competing hypotheses (ACH) is a structured analytic technique developed by Richards Heuer at the CIA to reduce the effect of cognitive biases — particularly mirror-imaging, confirmation bias, and anchoring — on intelligence assessment. ACH forces analysts to evaluate evidence against multiple hypotheses simultaneously rather than building a case for the most likely explanation.
The method proceeds in steps: (1) identify all reasonable hypotheses, not just the leading candidate; (2) list all significant evidence and arguments for and against each hypothesis; (3) construct a matrix with hypotheses as columns and evidence as rows; (4) assess the diagnostic value of each piece of evidence — whether it distinguishes between hypotheses rather than merely being consistent with the favored one; (5) refine the matrix by removing evidence that is consistent with all hypotheses (and therefore diagnostically useless); (6) draw tentative conclusions based on which hypotheses are least contradicted by the evidence; (7) identify what information, if collected, would most effectively discriminate among remaining hypotheses.
The key insight of ACH is that the most useful evidence is not that which confirms the leading hypothesis but that which disconfirms alternatives. Analysts are naturally drawn to confirmatory evidence; ACH redirects attention to disconfirmation, which is more informative under adversarial conditions where the adversary may be deliberately feeding confirmatory signals through denial and deception.
ACH’s limitations are well documented. It depends on the analyst generating the correct set of hypotheses — if the true explanation isn’t among the candidates, the method cannot find it. It also requires honest assessment of evidence against favored hypotheses, which institutional and psychological pressures can undermine.
The hypothesis-generation problem — ACH’s most critical known limitation — takes on a new dimension in the context of synthetic adversarial ecologies. Classically, the risk is that the analyst fails to include the correct adversary or the correct motive among the candidates. Against autonomous computational systems, the risk is more fundamental: the hypothesis space must include the possibility that no intentional actor exists at all. Behavior that ACH would normally attribute to one of several adversaries may instead be emergent, stochastic, or the byproduct of evolutionary dynamics. Agents of Angletonian Wilding argues that ACH’s disconfirmation logic remains useful when extended to this question — using the matrix not to determine which adversary acted but to determine whether the concept of “adversary” applies to the observed phenomena in the first place.
Related terms
- Mirror-imaging — the cognitive bias ACH is designed to mitigate
- Red teaming — a complementary technique that simulates adversary decision-making
- Denial and deception — adversary activity that makes ACH’s emphasis on disconfirmation critical