Differential Diagnosis Protocol¶

Diagnostic protocol — instantiates Observational Equivalence Resolution

Holds the full set of candidate explanations open and eliminates them one at a time against discriminating signs, refusing to close on the vivid front-runner until its rivals are actively ruled out.

Version: v1 · 2026-08-24 · History
Mechanism #: 2748
Type: Diagnostic Protocol
Form family: Protocol, Workflow & Routine
Solution family: Representation & Modeling
Problem family: Uncertainty, Evidence & Inference Failure
Problem subfamily: Explanatory Hypothesis, Pattern & Case Reasoning
Origin domain: Medicine & Healthcare
Instantiates: Observational Equivalence Resolution

The Differential Diagnosis Protocol is the enumerate-and-exclude mechanism. Faced with a shared observation that several causes could produce, it refuses to reason forward from the most obvious story and instead lays out the whole field of candidates first — deliberately including the unlikely-but-dangerous ones — and then works the list down by matching each candidate to a sign, test, or exclusion that would separate it from the rest. Its defining move is completeness before commitment: a diagnosis is not "the explanation that fits" but "the one candidate still standing after its rivals have been actively excluded." That discipline is what distinguishes it from a mechanism that zeroes in on two rival stories and contrasts their predictions — this protocol keeps the entire equivalence class visible until the evidence thins it.

Example¶

A 34-year-old arrives at an emergency department with acute pain in the right lower quadrant. The pain is the shared observation, and on its own it points nowhere in particular: appendicitis, ovarian torsion, an ectopic pregnancy, a ureteric stone, mesenteric adenitis, and a Crohn's flare can all present this way. The protocol's first act is to write that whole list down — not to pick the most probable, but to make the equivalence class explicit, with the can't-miss emergencies (ectopic, torsion) held on it even though they are statistically less likely.

Then each candidate is paired with a discriminating observable. A β-hCG test would come back positive for an ectopic pregnancy and negative for the others; a pelvic ultrasound would show the twisted ovary of torsion; urinalysis would flag the blood of a stone; imaging would show the inflamed appendix. The tests are ordered by how decisively they separate the field and how dangerous the candidate is if missed. β-hCG returns negative — ectopic excluded. Ultrasound shows normal ovaries — torsion excluded. A CT confirms appendicitis while the earlier exclusions have cleared the dangerous rivals. Only now, with the class narrowed to one at the required level of confidence, does treatment commit to an appendectomy. Had the tests come back inconclusive, the honest output would not be a guess but a handoff to observation.

How it works¶

The protocol runs a fixed cycle. Enumerate the differential broadly, biasing toward inclusion so that a rare-but-lethal cause is never omitted merely because it is unlikely. Frame the equivalence class — state which candidates the presenting picture genuinely cannot yet separate, so they are carried together rather than silently dropped. Attach a discriminator to each candidate: the finding it predicts that its rivals do not, expressed as a confirmation or an exclusion. Order the workup by discriminating power weighted against danger-if-missed, so the most decisive and most consequential tests run first. Run to threshold — continue eliminating until one candidate is confirmed and its rivals excluded to the confidence the stakes demand; if the field will not narrow, report residual ambiguity rather than force a winner.

Tuning parameters¶

Differential breadth — how many candidates are admitted. A wide list protects against candidate omission but costs time and tests; a narrow one is fast but risks missing the true cause entirely.
Ordering rule — rank the workup by prior probability, or by danger-if-missed (going first for the test that would catch the lethal long shot). The trade is efficiency against safety.
Confirm-vs-exclude weighting — whether the protocol advances by ruling candidates in or ruling them out. Exclusion-first is safer under ambiguity; confirmation-first is faster when one candidate dominates.
Test sequencing — parallel (order everything at once) versus serial (let each result gate the next). Parallel is quick and costly; serial is cheap and slow.
Closure threshold — how thoroughly rivals must be excluded before the diagnosis is treated as settled — the dial between premature closure and endless testing.

When it helps, and when it misleads¶

Its strength is that it makes anchoring hard: because the dangerous rivals are written down at the start, the vivid front-runner cannot quietly become the answer without earning it. It converts a hunch into an auditable elimination.

Its failure modes live at the two ends of the list. If a candidate never enters the differential, no amount of testing will surface it — candidate omission produces a confident, complete-looking workup that is simply blind to the true cause. At the other end is premature closure: treating a finding that is merely compatible with the favored diagnosis as if it confirmed it, before the rivals were excluded. The classic corrective in diagnostic teaching is Sutton's law — go first for the test that would most decisively separate the field, rather than the one that merely confirms your favorite.^[n1] The guarding discipline is to keep every can't-miss rival explicitly live until a discriminating observation has ruled it out, and to treat an inconclusive workup as unresolved rather than as endorsement of the leading guess.

How it implements the components¶

The Differential Diagnosis Protocol realizes the enumeration-and-elimination face of the archetype — building the candidate field and thinning it against evidence:

candidate_explanation_set — the differential itself: the deliberately broad list of causes that could produce the shared observation, including dangerous long shots.
observational_equivalence_class — the explicit statement of which candidates the current findings cannot yet separate, so they are carried together rather than dropped.
discriminating_observable — the sign, test result, or exclusion attached to each candidate that it predicts and its rivals do not.
evidence_threshold — the confirm-and-exclude bar that decides when a candidate is settled or ruled out.

It does not manufacture a contrast by acting on the system (discriminating_test_design, harm_constraint) — that is Ablation or Perturbation Test; nor does it separate candidates by pairing their predictions over naturally occurring variation (counterfactual_prediction_pair), the move of its nearest twin, Causal Identification Probe, which isolates two rival stories rather than thinning an open field.

Instantiates: Observational Equivalence Resolution — this protocol is the archetype's exhaustive candidate-elimination engine.
Sibling mechanisms: Causal Identification Probe · Ablation or Perturbation Test · Forensic Discriminator · Frame-of-Reference Shift · Side-Channel Measurement · Decision Tree with Hold State · Controlled Disambiguation Test · Ambiguity Register

Editorial Notes¶

Form Classification¶

Form family: Protocol, Workflow & Routine

Rationale: Differential Diagnosis Protocol operates as a repeatable ordered procedure or handoff sequence that coordinates action because it holds the full set of candidate explanations open and eliminates them one at a time against discriminating signs, refusing to close on the vivid front-runner until its rivals are actively ruled out.

Independent corroboration: The frozen evidence defines Differential Diagnosis Protocol as 'Holds the full set of candidate explanations open and eliminates them one at a time against discriminating signs, refusing to close on the vivid front-runner until its rivals are actively ruled out', so its operative form is Protocol, Workflow & Routine.

Review outcome: Independent reviewer agreement; high confidence.

Origin Attribution¶

Primary origin: Medicine & Healthcare

Origin pattern: Single lineage

Present-day reach: Multi-domain

Rationale: Clinical diagnostic training cohered enumerate-and-exclude protocols that keep the whole candidate field visible until decisive tests thin it.