Decision-Process Review¶
Structured review — instantiates Metacognitive Monitoring Loop
Examines how one decision was reasoned — assumptions, alternatives, evidence, confidence — and grades the process quality separately from how the outcome happened to turn out.
Most reviews of a decision ask "did it work?" Decision-Process Review refuses that question and asks a different one: was it reasoned well? It takes a single, specific decision and reconstructs the machinery behind it — the assumptions taken as given, the alternatives that were and weren't considered, how hard the team searched for disconfirming evidence, whether the stated confidence was warranted — and grades that machinery on its own terms. Its defining commitment is the separation of process from outcome[1]: a good decision can be followed by a bad result and a reckless one by a lucky win, so judging the reasoning by the result ("resulting") corrupts every lesson you might draw. The output is not a verdict on the choice but a named diagnosis of the reasoning — where it was sound, and where a specific process failure crept in.
Example¶
A venture fund's investment committee revisits its decision, eighteen months earlier, to lead a round in a logistics startup that has since stalled. The tempting review is short: "the market turned, bad luck." The Decision-Process Review does not let them stop there. It pulls the original memo and walks the reasoning. The thesis rested on one assumption — that a key customer would expand — recorded as fact, never pressure-tested. Two partners had flagged a competing entrant; the notes show the concern was acknowledged and then never revisited. The base-rate on companies at that stage was in the deck but played no role in the debate. The confidence expressed ("high conviction") had no evidence behind it beyond the founder's charisma.
The review's product is not "we were wrong to invest." It is a labeled finding: the process failed by anchoring on an untested load-bearing assumption and by premature closure around a preferred narrative. Those are named, recurring failure modes the committee can now watch for — and, crucially, the same diagnosis would stand even if the startup had, by luck, succeeded. That is the point: the review grades the reasoning the fund can repeat, not the outcome it can't control.
How it works¶
- Fix the object as a single decision. The unit of review is one bounded choice with a reconstructable rationale, not a whole program or a person.
- Reconstruct the reasoning, not the story. Recover the actual assumptions, options, evidence search, and stated confidence from contemporaneous artifacts — memos, notes, models — before hindsight rewrites them.
- Hold outcome constant. Deliberately bracket how it turned out; ask whether the reasoning was sound given what was knowable at the time.
- Name the failure mode. Close with a specific, named process fault — anchoring, premature closure, shallow search, overconfidence — rather than a diffuse "could have been better."
Tuning parameters¶
- Independence — self-review, peer review, or an outside panel. Distance reduces defensiveness and blind spots but loses context the deciders held tacitly.
- Timing — run before the outcome is known (a pure process check) or after (richer, but must fight hindsight). Pre-outcome reviews are cleaner; post-outcome ones are more available.
- Scope of the reconstruction — a light walk-through or a full forensic rebuild from primary artifacts. Depth surfaces subtler faults at real cost in time.
- Blame stance — diagnostic (name the process fault) versus evaluative (grade the decider). A diagnostic stance protects the candor the review depends on; an evaluative one tends to end candor.
When it helps, and when it misleads¶
Its strength is that it makes reasoning quality auditable independent of luck, which is the only way an organization learns transferable lessons rather than superstitions. By insisting on a named failure mode, it also builds a shared vocabulary — "that's anchoring again" — that speeds future monitoring.
Its central failure is hindsight contamination: once the outcome is known, the reasoning is almost impossible to see as the deciders saw it, and a review can quietly grade the process for failing to predict what only became obvious afterward. A second failure is rationalization in reverse — a review run to justify or to blame, working backward from a conclusion. And because it examines people's thinking, it can curdle into a tribunal, at which point everyone starts defending rather than disclosing. The guarding discipline is to reconstruct from contemporaneous artifacts, keep the stance diagnostic rather than punitive, and — where possible — review the process before the result is in.
How it implements the components¶
process_monitor— the review is an observer trained on the reasoning process rather than the object-level result; that reorientation is its defining act.reasoning_failure_mode_label— its deliverable is a specific, named process fault (anchoring, premature closure, shallow search), which targets any future correction.
It stops at diagnosis. It does not reconstruct a whole completed effort to decide what future strategy should change (reflection_step, strategy_adjustment) — that forward, learning-oriented reconstruction is [Reasoning Retrospective], its nearest twin; the two differ in that this review audits and labels one decision's reasoning while the retrospective re-plans the next one. Nor does it run as live developmental dialogue during the work (external_feedback_source) — that is [Metacognitive Coaching] — or calibrate the confidence signal itself (confidence_check), which is [Confidence Rating Rubric].
Related¶
- Instantiates: Metacognitive Monitoring Loop — it monitors a decision's reasoning process after the fact and names where it failed.
- Sibling mechanisms: Reasoning Retrospective · Metacognitive Coaching · Confidence Rating Rubric · Learning Journal · Strategy Checklist · Reflection Prompt · Think-Aloud Protocol
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: Decision-Process Review operates as a bounded evaluation of existing evidence or work that produces a finding or disposition because it examines how one decision was reasoned — assumptions, alternatives, evidence, confidence — and grades the process quality separately from how the outcome happened to turn out.
Independent corroboration: The frozen evidence defines Decision-Process Review as 'Examines how one decision was reasoned — assumptions, alternatives, evidence, confidence — and grades the process quality separately from how the outcome happened to turn out', so its operative form is Assessment, Review & Assurance.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Psychology
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Judgment and decision-making research cohered process review that grades assumptions, alternatives, evidence search, and calibration separately from outcomes to avoid outcome bias and resulting.
Related originating lineages:
- Organizational & Management Science — Decision journals and post-decision reviews operationalized the reconstruction inside organizations.
Review resolution: Judgment and decision-making research cohered process review that grades assumptions, alternatives, evidence search, and calibration separately from outcomes to avoid outcome bias and resulting.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; high confidence.
References¶
[1] Baron, J., & Hershey, J. C. "Outcome Bias in Decision Evaluation". Journal of Personality and Social Psychology 54(4), 569–579 (1988). Separates decision quality from outcome quality and shows that evaluators judge the same decision differently after good and bad outcomes. registry ↩