Decision Closure Criteria¶

Closure protocol — instantiates Regress Termination Rule

Converts abstract 'enough support' into a concrete, decision-class checklist that says when a specific decision may be closed and acted on.

"Do we have enough to decide?" is a question that can be asked forever. Decision Closure Criteria answer it once, in advance, by turning the vague notion of sufficient support into a concrete checklist attached to a class of decision: when these named conditions are satisfied, this decision is closed and acted on — full stop. What makes it THIS mechanism is that it governs the readiness of the whole decision (evidence, approvals, risk checks, and all), calibrated to the decision's stakes — not the quality of any single input. It is the general gate that says "we may now stop and act," which is exactly why it is not the same as grading the evidence or stopping the research.

Example¶

A hiring committee keeps circling — should we do one more panel? another reference? — and good candidates go cold while it deliberates. So it defines closure criteria for the "senior engineer" decision class, before seeing any candidate: at least three independent interviewers rate each core competency at or above the bar; one work-sample is completed and reviewed; two references are checked; and no red flag is raised by more than one interviewer without resolution. The moment a candidate's file satisfies every line, the decision closes — hire or no-hire — and the reflex to gather "just a little more signal" no longer has standing. Because the criteria are set by class, a three-month contractor rides a lighter checklist and a VP a heavier one; the same committee stops at different points for different stakes, on purpose.

How it works¶

Name the decision and its class. Fix exactly which choice is being held open and which stakes bucket it falls in; the checklist is indexed to the class, not improvised per case.
Translate "enough" into checks. Restate sufficiency as observable, countable conditions a bystander could verify, so closure is a fact rather than a feeling.
Set the bar by stakes and reversibility. A reversible, low-stakes call gets a light checklist; an irreversible or rights-affecting one gets a strict one.
Close on satisfaction. When the checks are met, the decision is closed and the demand for further support loses standing — the regress terminates at the checklist.

Tuning parameters¶

Class granularity — one checklist for everything versus a checklist per decision class. Finer classes calibrate better but multiply the rules to maintain.
Strictness by stakes — how high the bar sits for the class. Higher bars reduce bad closures but raise the cost and latency of every good one.
Check observability — how objective and countable each criterion is. Objective checks resist gaming; softer, judgment-based ones capture nuance but invite drift.
Combination logic — whether all criteria must pass or a weighted subset suffices. Strict conjunction is safe but brittle; weighting is flexible but easier to fudge.

When it helps, and when it misleads¶

Its strength is converting an open-ended "are we sure yet?" into a rule that can actually be met, killing the "just one more" churn and making closure a decision the group can defend. It scales especially well across repeated decisions of the same kind, where per-case deliberation is pure waste.

Its characteristic failure is checklist theatre: satisfying the letter of the criteria while the substance is plainly wrong — every box ticked on a hire everyone privately doubts — because a checklist rewards form over judgment. The mirror failure is setting the bar low enough to feel productive, so closure arrives before the decision is real. And like any closure rule it can be run backwards: criteria chosen because the favored option already meets them. This is deliberately satisficing rather than optimizing — good-enough-to-act, not best-possible^[1] — so the discipline is to fix the criteria before the options are known, calibrate the bar to genuine stakes, and keep at least one criterion that can register a substantive objection the boxes would otherwise hide.

How it implements the components¶

claim_under_justification — it pins the exact decision being held open, so the checklist grounds a specific choice rather than a moving target.
stopping_criterion — the checklist is the operational stopping criterion: a set of practical, verifiable conditions that count as enough for this decision class.
sufficiency_threshold — it calibrates the strength of that criterion to the decision's stakes and reversibility, tightening or loosening the whole gate.

It does not calibrate the quality and diversity of the *evidence itself (sufficiency_threshold for evidence → Evidence Sufficiency Rubric), name the base premises being reasoned from (foundational_assumption → First-Principles Statement), or list what would reopen the closed decision (reopening_condition → Review Trigger Register).*

Instantiates: Regress Termination Rule — it is the general decision-readiness gate that terminates the "are we sure enough?" regress.
Consumes: Evidence Sufficiency Rubric — where a criterion is evidentiary, the rubric's verdict feeds in as one check.
Sibling mechanisms: Evidence Sufficiency Rubric · Research Stopping Rule · Burden-of-Proof Rule · Axiom Set · First-Principles Statement · Five Whys with Stop Rule · Governance Authority Chain · Review Trigger Register · Timeboxed Inquiry Record · Assumption Log

Notes¶

Closure Criteria gate a decision; a Research Stopping Rule gates data collection. They often chain — you stop gathering, then check whether what you have satisfies the closure checklist — but they answer different questions and should not be merged, or a team will conflate "we ran out of things to look at" with "we have enough to decide."

References¶

[1] Herbert Simon's satisficing — accepting the first option that clears an aspiration level rather than searching for the optimum — is the logic these criteria encode: closure is defined as good-enough-for-this-purpose, not provably best. ↩