Reduces confirmation bias by simultaneously testing every plausible explanation.
Purpose
Instead of looking for evidence that confirms a preferred hypothesis, ACH lists every competing hypothesis, scores each piece of evidence against each, and seeks the hypothesis with the fewest contradictions.
Body
ACH was formalized by Richards Heuer (CIA) as an antidote to confirmation bias. The process has seven steps: (1) list hypotheses; (2) list evidence and arguments; (3) build the matrix; (4) refine — drop non-diagnostic evidence; (5) draw tentative conclusions; (6) sensitivity-test the conclusion against critical assumptions; (7) report conclusions with calibration. olivLaw exposes ACH as an interactive board in Intel Hub (table `mirofish_ach`).
Core concepts
- Competing hypotheses
- Hypothesis × evidence matrix
- Diagnosticity (how discriminating an item of evidence is)
- Elimination by contradiction
- Final confidence calibration
UI / implementation ideas
- Editable matrix with consistency codes (++ / + / 0 / − / −−)
- Per-hypothesis contradiction visualization
- Side-sheet per evidence with source and confidence
- Final per-hypothesis score and ranking
Related
I&W — Indicators & Warning
Early detection of weak signals before strategic events unfold.
Premortem — assumed-failure analysis
Assume the plan has failed, then identify the causes before it begins.
Red teaming — adversarial challenge
Attack your own assumptions and forecasts from the adversary's perspective.
Confidence calibration
"70% probability" should actually happen in 70% of cases.