Differential Diagnosis List¶
Working list — instantiates Alternative Hypothesis Generation
Enumerates a bounded, ranked list of candidate explanations for the same findings — including the unlikely-but-dangerous ones — then rules each in or out as evidence arrives, so the answer is the survivor of a field rather than the first guess.
A Differential Diagnosis List is the working roster of candidate explanations that could account for the same set of observations, held open and worked one by one until evidence narrows it to a winner. Borrowed from clinical medicine, its distinguishing move among its siblings is that it demotes the leading hypothesis to a peer entry on a ranked list and then runs a disciplined rule-in / rule-out process against that list: no candidate is the answer until the others have been positively excluded. Where a pure generator only opens the field and a matrix only maps which evidence separates rivals, the differential is the artifact that carries the bounded set and retires its members — its whole value is in being worked, not merely written.
Example¶
A patient arrives at an emergency department with sudden chest pain. The reflex read — "probably a pulled muscle or anxiety" — is comfortable and often right, which is exactly the danger. Instead of accepting it, the clinician writes a differential: musculoskeletal strain, acid reflux, panic, pericarditis, heart attack, pulmonary embolism, aortic dissection. The list is ranked on two axes at once — how likely each is, and how deadly it is to miss — so the rare aortic dissection stays on the list despite low probability because it kills fast. Then the list is worked: an ECG and troponin address the cardiac entries, a D-dimer and the clinical picture speak to the clot, breath sounds and history chip at the rest. As each result lands, an entry is struck only when it is genuinely excluded, and the surviving candidate — not the first one that came to mind — becomes the diagnosis.
How it works¶
The list is generated for a single presentation, then filtered and worked:
- Enumerate the candidates that could produce these findings, deliberately including the leading hunch as just one line item.
- Apply a plausibility floor and relevance filter — drop what is physically impossible or irrelevant to this case, but keep the unlikely-but-catastrophic ("must not miss") entries that a naïve probability cut would delete.
- Rank by combined likelihood and cost-of-missing.
- Rule in / rule out as evidence arrives: each entry stays alive until an observation positively excludes it, and a struck entry can be reopened if a new finding contradicts the closure.
The signature is that last discipline — the list refuses to collapse to one answer until the field has actually been narrowed by evidence.
Tuning parameters¶
- List breadth — how many candidates are admitted. A wider list catches the rare "zebra" but multiplies tests and effort; a narrow one is fast but blind to the unusual.
- Plausibility-floor height — how aggressively improbable entries are pruned. Set it high and you move fast; set it too high and you delete the rare-but-lethal.
- Must-not-miss weighting — how much a dangerous-but-unlikely rival is protected from the probability cut. More protection means more over-testing traded against fewer catastrophic misses.
- Exclusion standard — how strong a negative result must be to strike an entry. A strict standard resists premature closure but keeps the list open longer.
- Re-open trigger — whether, and on what, a struck entry comes back.
When it helps, and when it misleads¶
Its strength is that it forces the reflexive first answer to compete, and its must-not-miss rule catches the rare killer that a probability-only shortcut would drop. It turns "I think it's X" into "X is leading a field of six, and here is what would take it down."
Its failure modes are the classic diagnostic-error pair. Anchoring fixes attention on the first or most vivid entry, and premature closure declares that entry the winner before the rest of the field is truly excluded — the single most common route to a missed diagnosis.[1] The other misuse is a list that is written but never worked: entries enumerated for form's sake and then abandoned, so the artifact decorates a decision already made. The discipline that guards against both is requiring a positive exclusion to strike each entry and keeping a re-open trigger live, so the list stays an instrument rather than a ritual.
How it implements the components¶
focal_claim_or_leading_hypothesis— the presenting hunch is entered as one ranked line, not privileged above the rest.alternative_hypothesis_set— the differential is the bounded, ranked set of rivals for the same findings.plausibility_floor_and_relevance_filter— the prune-but-protect step: drop the impossible and irrelevant while keeping the must-not-miss.rival_retention_or_elimination_rule— the rule-in / rule-out discipline is its defining machinery; entries live until positively excluded.
It does not map *which observation discriminates each pair of rivals (discriminating_evidence_map) — that is Discriminating Test Matrix — nor does it issue the final calibrated confidence in the survivor (confidence_and_warrant_label), which is Red-Team Rival Explanation Review's output.*
Related¶
- Instantiates: Alternative Hypothesis Generation — it holds the rival field and narrows it to a warranted answer.
- Sibling mechanisms: Why Else Could This Be True? Prompt · Discriminating Test Matrix · Counter-Narrative Probe · Base-Rate Alternative Prompt · Red-Team Rival Explanation Review
Editorial Notes¶
Form Classification¶
Form family: Representation, Specification & Plan
Rationale: Differential Diagnosis List operates as a non-executable information artifact that externalizes static or prospective structure because it enumerates a bounded, ranked list of candidate explanations for the same findings — including the unlikely-but-dangerous ones — then rules each in or out as evidence arrives, so the answer is the survivor of a field rather than the first guess.
Independent corroboration: The frozen evidence defines Differential Diagnosis List as 'Enumerates a bounded, ranked list of candidate explanations for the same findings — including the unlikely-but-dangerous ones — then rules each in or out as evidence arrives, so the answer is the survivor of a field rather than the first guess', so its operative form is Representation, Specification & Plan.
Nearest alternative: Decision, Gate & Allocation — The ranked differential is a persistent case artifact that keeps candidates visible while later evidence determines disposition.
Review outcome: Independent reviewer agreement; medium confidence.
Origin Attribution¶
Primary origin: Medicine & Healthcare
Origin pattern: Single lineage
Present-day reach: Multi-domain
Rationale: Clinical practice cohered the ranked differential list, including unlikely-but-dangerous conditions, as a working artifact whose candidates are ruled in or out over time.
Review outcome: Independent reviewer agreement; high confidence.
Notes¶
The demotion of the leading hypothesis to a peer entry is the anti-anchoring core, and it is easy to lose. A differential written after the clinician has silently decided, with the favored answer at the top and the rest as garnish, has the shape of the mechanism without its function. The list only does its work when the favored candidate can genuinely lose to a line beneath it.
References¶
[1] Croskerry, P. "The Importance of Cognitive Errors in Diagnosis and Strategies to Minimize Them". Academic Medicine 78(8), 775–780 (2003). Defines anchoring as early fixation on salient initial features and premature closure as accepting a diagnosis before full verification. registry ↩