A single decision audit is examination. A record of decisions is something else — a mirror that compounds. Patterns you cannot see from inside one choice become visible across ten: the frames you repeat, the incentives you never name, the confidence that stays flat whether the stake is a hire or a relocation.
DAUDIT remembers because judgment without memory is amnesia with confidence. You make the same category of mistake, in a new costume, every few years — and call it a new situation.
Pattern Echo
Pattern Echo runs when you audit a new decision. It looks for structural rhymes in your prior audits — not to tell you “you always do X,” but to ask whether the current frame rhymes with a prior frame that did not end where you expected.
Examples (generalized):
- You are optimizing terms for a partnership you have not decided to enter — echo of a prior co-founder/equity audit
- You are comparing facilities before asking whether moving is on the table — echo of a prior care-decision audit
- You are adding to a position because the price dropped — echo of a prior thesis-that-was-not-a-thesis audit
Echo is not destiny. It is a prompt to examine faster — the pre-mortem before the pre-mortem.
Pattern Echo is forming, not proven at population scale. It works best when you have multiple real audits in the system — which is why the instrument rewards honest use over time, not one-off performance.
How you decide
How you decide is a structured map of decision tendencies across lanes — how you relate to risk, time, conflict, and evidence over time. It is built from your audit history, not from a quiz that asks whether you consider yourself “analytical.”
What it is:
- A descriptive model of how you have decided
- Updated as new audits resolve and outcomes arrive
- A input to calibration — connecting what you believed to what happened
What it is not:
- A fixed personality type
- A score to optimize for social sharing
- A prediction that you will fail — it is a map, not a verdict
How you decide lanes are honest about uncertainty. Early audits produce provisional shapes. The map sharpens with record — the same way calibration only works if you keep score.
The Decision Ledger
Underneath Echo and How you decide is the Ledger — your private record of decisions examined, confidence stated, outcomes logged. This is the infrastructure most people never build for themselves: a place where “I was 80% sure” meets “here is what happened” without the story that protects ego.
The Ledger is what makes DAUDIT different from a chat session. Chat forgets. Forgetting is comfortable. It is also why people repeat frames.
Matthews’s recurring discipline is the same spine in plain language: a sound decision can produce a bad result, and a careless one can produce a good result — so outcome quality is not decision quality. Without a record, success rewrites memory. The Ledger is how you refuse that rewrite.
Same insight. Two lifetimes.
Frame, probability, unasked question — at the time of deciding
Outcome meets the record without protecting ego
Echo asks whether the new frame rhymes with an old miss
What compounds — and what does not
Compounds:
- Calibration accuracy over dozens of scored predictions
- Pattern recognition across domains (capital rhymes with life more often than people admit)
- Faster examination on new decisions because Echo names the rhymes early
Does not compound (yet / honestly):
- Population-level claims about “users who audit X do Y” — we do not have the sample or the ethics to claim that
- Automatic “right answer” prediction — the instrument refuses that job by design
- Replacement for domain expertise — auditing a biotech investment does not teach you biology
We would rather say forming where the science is early than proven where it is not. That is the same honesty rule the audit applies to your decisions.
What is proven outside DAUDIT is narrower and useful: keeping score improves judgment. In the Good Judgment Project's geopolitical forecasting tournaments, a short cognitive-debiasing training module — including calibration, base rates, and tracking probabilistic forecasts — improved Brier scores by roughly 6–12% relative to controls across tournament years. Recorded probabilities beat remembered conviction. That is the ledger's reason to exist — not as a claim that DAUDIT users become “superforecasters,” but as evidence that memory-plus-score is not a gimmick.
Forecast training + scorekeeping vs. control (Good Judgment Project)
Source: Chang, Chen, Mellers & Tetlock — Developing expert political judgment (Judgment and Decision Making, 2016) — Approximate relative Brier-score improvement from probability / scenario training vs. controls, by tournament year. Lower Brier is better; bars show reported % improvement.DAUDIT does not claim these gains as product metrics. The chart is why the Ledger is not optional for anyone who wants calibration — feelings do not self-correct; records do.
Why this matters for the skeptical reader
If you only need one examination, audit once and take the close. Memory is optional.
If you make consequential decisions regularly — and you do, whether or not you call them “strategic” — memory is not a feature. It is the difference between improving judgment and performing improvement.
Read What a decision audit actually is for the session itself. Read [Confidence is not evidence](/blog/confidence-is-not-evidence) for why the record must include scores, not feelings.
The instrument is young. The method is old. Memory is how the method learns you — with your permission, in private, without selling the story back to you as certainty.