Clarity Matrix · Analysis of Competing Hypotheses

A structured way to weigh competing explanations — by trying to disprove them, not confirm them. Runs entirely in your browser; nothing leaves this page.

The decision / question

Competing hypotheses — list them ALL, even unlikely ones. A hypothesis not on the table can't win.

Evidence & arguments — grade each source, then mark how it sits against each hypothesis. Click a cell to cycle.

cell: · n/a   C consistent   I inconsistent   II strongly inconsistent
Source grade = Admiralty code: reliability A (best) → F (can't judge) × credibility 1 (confirmed) → 6 (can't judge). A weak grade down-weights that row.

Result — the LEAST-inconsistent hypothesis leads. Consistent evidence doesn't prove anything; only inconsistency eliminates.

Assessment — state likelihood and confidence SEPARATELY. They are two different things (ICD 203).

Assign a likelihood and a confidence above to generate a properly-worded assessment.
How this works — and the honest limits of it

The method is Richards Heuer's Analysis of Competing Hypotheses (CIA, 1970s). You enumerate every plausible explanation, line up the evidence, and rate each piece of evidence as consistent or inconsistent with each hypothesis. You then pick the hypothesis with the least inconsistent evidence — because evidence that's consistent with a hypothesis is usually consistent with the rival hypotheses too, so it can't single one out. Only an inconsistency eliminates.

Why "diagnostic value" matters. Evidence consistent with every hypothesis tells you nothing, no matter how solid or vivid it is. This tool flags those rows as dead weight so you stop giving them credit.

The Curveball alarm. If your conclusion flips when you remove one weak or uncorroborated source, you don't have a conclusion — you have a single point of failure. That exact mistake (one fabricating source, "Curveball," treated as fact) drove the 2002 Iraq-WMD estimate. The tool watches for it.

Honest limits. Controlled studies do not show ACH reliably beats unaided judgment at getting the right answer. What it reliably does is make your reasoning transparent and auditable — it forces the full hypothesis set onto the table, separates evidence from assumption, and exposes weak linchpins. The part of intelligence work that measurably improves accuracy is calibration: stating probabilities and keeping score over time. Use this to structure the argument; don't mistake a tidy matrix for a correct one.

Built from scratch, no dependencies, no tracking. Your work is saved only in this browser.