Decision Checklist¶
Checklist — instantiates Bias-Specific Decision Audit
A short, decision-specific list of the few bias checks worth running here, phrased as prompts a reviewer answers before closing — kept deliberately brief so it actually gets used.
A Decision Checklist is the short list of bias checks a reviewer runs through before closing a specific kind of decision — phrased as questions to answer, not steps to log. Its defining property is brevity under a load budget: it is not the full audit and not the record of one, but the handful of prompts distilled from a decision type's known vulnerabilities, kept short enough that a busy person will actually run it at the moment of choice. A checklist with forty items is a poster no one reads; a checklist with four is a pause that catches the predictable mistake. It carries the audit's conclusions forward into the decision — "did you look for the disconfirming source? did you consider the alternative you dismissed first?" — where a review form carries the same review backward into a record.
Example¶
A newsroom's standards editor keeps a five-line checklist taped beside every decision to publish a sensitive investigative story. Before a piece goes live, the editor answers: Have we sought the strongest source that would contradict the story's thesis? Have we given the accused a real chance to respond? Are we relying on a single source for any load-bearing claim? Would we frame a friendly subject's actions the same way? Is the timing driven by the story or by the scoop?
On a story about a local official, the checklist stops the editor cold on line one — the reporting has three sources who all agree and none who push back, a confirmation pattern the deadline had obscured. The story holds for a day while a disconfirming source is sought. The checklist did not analyze the decision or file a record; it simply made the reporter and editor pause on the two questions most likely to matter before the story became irreversible.
How it works¶
- Distill, then cut. Start from the decision type's vulnerability map and reduce it to the few checks whose absence would actually change the outcome; the discipline is in what you leave out.
- Phrase as prompts, tied to action. Each item is a question with a clear consequence, not a topic. A prompt the reviewer can answer "no" to — and must then do something about — is a check; a prompt that invites "yes, of course" is decoration.
- Fire before closure. The checklist is run at the last reversible moment, so a "no" can still change the decision rather than merely annotate it.
- Keep it short on purpose. Every item added competes for the attention the checklist depends on; the list earns its length only by staying near its load budget.
Tuning parameters¶
- Length — the master dial. Short lists get used and cover less; long lists cover more and get skipped. Bias toward short: a thorough checklist no one runs beats nothing only in theory.
- Specificity — generic ("watch for bias") versus decision-specific ("is any claim single-sourced?"). Specific items catch real errors; generic ones are the checklist-theater failure mode in miniature.
- Answer format — open reflection versus forced yes/no/N-A. Forced answers make skipping visible; open ones invite hand-waving.
- Ownership — who is accountable for each answer. An item no one owns is an item no one runs.
When it helps, and when it misleads¶
Its strength is friction at the right moment for almost no cost: a good checklist converts a decision type's hard-won lessons into a thirty-second pause that reliably catches the predictable slip, which is why short checklists cut error in domains from surgery to aviation.[n1]
Its failure mode is the generic all-bias checklist — the long, universal list of cognitive biases that feels rigorous and changes nothing, because a reviewer facing forty abstract prompts pattern-matches "yes" down the page. A checklist also goes stale silently: the pathways that mattered last year may not be this year's, and an unrevised list keeps prompting for the wrong risks. The discipline that guards against this is to keep the list short, specific, and tied to a consequence, and to prune it whenever an item stops catching anything — the checklist's power is entirely in its brevity and fit, and both decay if it is left alone.
How it implements the components¶
targeted_bias_check— each item is a specific check aimed at a mapped vulnerability, expressed as an answerable prompt.audit_load_budget— brevity is the mechanism, not a side effect: the checklist exists to hold the review to the few checks the decision's stakes justify.
It does not build the vulnerability map its items are distilled from (bias_vulnerability_map — that is Bias Audit), and it does not capture what was found as a durable record (evidence_and_process_trace — that is Structured Review Form); a checklist prompts forward, the form records backward.
Related¶
- Instantiates: Bias-Specific Decision Audit — the Decision Checklist is the audit's lightweight, point-of-use edge, carrying its conclusions into the moment of choice.
- Consumes: Bias Audit — the checklist is the short distillation of that audit's vulnerability map; without the map behind it, the list is guesswork.
- Sibling mechanisms: Bias Audit · Structured Review Form · Hiring Review Rubric · Blind or Masked Review · Independent Estimation · Diagnostic Debiasing Check · Decision Log · Reference-Class Forecasting
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: At the last reversible moment a reviewer applies a short set of consequential bias prompts and must act on each failed check before closure, so its operative form is decision-quality assessment.
Nearest alternative: Representation, Specification & Plan — The checklist carries the criteria, but the mechanism depends on evaluating the live decision rather than maintaining a passive reference list.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Medicine & Healthcare
Origin pattern: Convergent development
Present-day reach: Multi-domain
Rationale: Clinical safety practice cohered short, high-yield checklists used immediately before consequential action to catch predictable omissions under workload.
Related originating lineages:
- Aviation & Aeronautics — Pre-flight and cockpit checklists provide an earlier safety-critical lineage for brief standardized verification.
- Psychology — Debiasing research supplied decision-specific prompts for disconfirmation, alternative generation, and bias interruption.
Review resolution: Clinical safety practice cohered short, high-yield checklists used immediately before consequential action to catch predictable omissions under workload.
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.
Notes¶
[n1] The clinical case for short, targeted checklists is most associated with Atul Gawande's The Checklist Manifesto and Peter Pronovost's central-line-infection checklist in intensive care — both turn on a short list of high-yield items, not an exhaustive one, which is exactly the property a bias checklist must preserve. ↩