Decision Threshold Rule¶
Procedure — instantiates Hypothesis Testing Frame
Operationalizes the evidence threshold as a cut point, burden, gate, or standard that changes action status.
A Decision Threshold Rule is the mechanism that turns a continuous score into a discrete action by drawing a line: below it, one status; at or above it, another. It is deliberately indifferent to how the score was produced — its entire concern is where the cut belongs and what flips when evidence crosses it. Its defining move is to place that line by reasoning from the asymmetric cost of the two mistakes — a wrongful yes versus a wrongful no — rather than by convention or convenience. What makes this THIS mechanism is that it isolates the threshold itself as the object of design: not the test that generates the number, not the interpretation of what the number means, but the single tunable line where action status changes.
Example¶
A consumer lender scores each applicant's probability of default from 0 to 100 and must convert that continuous score into a binary act: approve or decline. The score is given; the question is where to put the approval cut. The lender reasons from error costs. A wrongful approval (lend to someone who defaults) costs the unpaid principal; a wrongful decline (refuse someone who would have repaid) costs the foregone interest margin and the customer. Because a default is far more expensive than a lost good customer, the cost of the two errors is lopsided, and the cut is set accordingly.
Moving the cut up — demanding a stronger score to approve — declines more marginal applicants, cutting defaults but also turning away good borrowers; moving it down does the reverse. There is no free setting: every position trades one error against the other, a tradeoff the lender reads off a sensitivity-versus-specificity curve.[n1] They land on a cut that holds expected default losses within appetite, and write the rule plainly: score ≥ 62 approves, below declines, and the cut does not move for an individual applicant after the score is in.
How it works¶
- Take the score as given. The rule consumes an evidence signal from elsewhere; producing or validating it is not its job.
- Price the two errors. Estimate the cost of a false accept and a false reject; their ratio, and the base rate, drive where the cut belongs.
- Place the cut. Choose the line that best balances the two error costs, reading the tradeoff off the sensitivity/specificity curve.
- Bind the action. Specify exactly what status changes on each side and freeze the cut so it cannot be nudged case-by-case after the score is seen.
Tuning parameters¶
- Cut location — the master dial. Raising it suppresses false accepts at the price of more false rejects; lowering it does the opposite.
- Error-cost ratio — how much worse one mistake is than the other; the whole justification for an off-center cut lives here.
- Number of bands — a single cut, or multiple (approve / refer / decline). More bands add nuance but complexity and more lines to defend.
- Hysteresis — whether crossing back requires re-clearing a margin, to stop borderline cases flapping between statuses.
When it helps, and when it misleads¶
Its strength is making the action rule explicit, auditable, and stable: everyone can see where the line is, why it sits there, and that it will not move for a favored case. Anchoring the line to error costs is what keeps a threshold from being arbitrary or merely conventional.
Its failure mode is false confidence in the cut's exactness — a score of 61 and 63 are treated as categorically different when they are barely distinguishable, and a line placed on stale cost estimates keeps firing long after the costs shifted. The classic misuse is post-hoc thresholding: moving the cut after seeing where a particular case fell so the desired action comes out. The guarding discipline is to set and record the cut from error costs before the decisive evidence is in, revisit it on a schedule as costs and base rates change, and treat scores near the line as genuinely uncertain rather than crisply sorted.
How it implements the components¶
evidence_threshold— the cut point itself: the mechanism's entire reason for being.decision_rule— the mapping from which side of the cut the evidence falls to the action status that results.error_cost_profile— the asymmetric cost of false accepts versus false rejects that justifies where the cut is placed.
It is agnostic to how the score arrives: it does not implement test_evidence or power_or_sensitivity_check — the sampling and sensitivity of the underlying measurement belong to Null Hypothesis Significance Test. This rule supplies only the line and what happens at it.
Related¶
- Instantiates: Hypothesis Testing Frame — the procedural core that turns any evidence signal into a status change.
- Sibling mechanisms: Null Hypothesis Significance Test · A/B Test Interpretation Protocol · Legal Burden-of-Proof Analog · Inspection Pass/Fail Test
Editorial Notes¶
Form Classification¶
Form family: Rule, Policy & Commitment
Rationale: Decision Threshold Rule operates as a standing rule, threshold, contractual commitment, or policy constraint governing future conduct because it operationalizes the evidence threshold as a cut point, burden, gate, or standard that changes action status.
Independent corroboration: The frozen evidence defines Decision Threshold Rule as 'Operationalizes the evidence threshold as a cut point, burden, gate, or standard that changes action status', so its operative form is Rule, Policy & Commitment.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Statistics & Experimental Design
Origin pattern: Single lineage
Present-day reach: Universal
Rationale: Statistical decision theory and hypothesis testing cohered action thresholds chosen by trading false-positive against false-negative loss on a continuous score.
Related originating lineages:
- Law & Governance — Legal burdens and standards of proof independently institutionalized thresholds that change an action's permissible status.
- Medicine & Healthcare — Diagnostic testing operationalized sensitivity, specificity, ROC curves, and clinically consequential cut points.
Review resolution: Statistical decision theory and hypothesis testing cohered action thresholds chosen by trading false-positive against false-negative loss on a continuous score.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
The threshold and the test that feeds it are deliberately separate mechanisms. That separation is what lets an organization improve its scoring model without re-litigating where the action line sits, and re-tune the line when costs shift without touching the model.
[n1] Any placement of a cut point on a continuous score trades sensitivity (catching true positives) against specificity (avoiding false positives); the receiver-operating-characteristic (ROC) curve plots this tradeoff across all possible cut locations. Choosing a threshold is choosing one point on that curve, and there is no location that improves both at once. ↩