Assumption Audit Worksheet¶
Checklist — instantiates Correspondence Violation Detection and Theory Refinement
Enumerates, one row per premise, which assumptions must hold for two formulations to correspond and marks which of them fails in the regime where agreement broke.
An Assumption Audit Worksheet treats a correspondence violation as a question about premises, not about outputs. It is a structured checklist — one row per assumption — listing every condition that must hold for the new and old formulations to be legitimately comparable, and marking, for the regime where agreement broke, which conditions still hold, which are violated, and which are unknown. Its defining move is that it never asks "why is the answer wrong?"; it asks "which of the things we assumed to be true stopped being true here?" It deliberately folds the mundane commensurability premises — same units, same denominators, same schema, same operational definitions — into the same ledger as the deep theoretical ones, because the archetype's most common failure is a translation mismatch masquerading as a theory failure. The worksheet localizes the broken premise; it does not size the divergence or propose the repair.
Example¶
An engineer models a cantilever bracket with a new nonlinear finite-element solver and expects it to match the closed-form Euler–Bernoulli prediction for tip deflection, δ = FL³/3EI, at small loads. At a moderate load the FEA result runs about 12% higher than the formula — an apparent violation. The Assumption Audit Worksheet lays the premises out in rows: small-strain (deflection ≪ length), linear-elastic material, plane sections remain plane, shear deformation negligible, fully-fixed root boundary, and the plain-commensurability rows — consistent units (N, mm), same Young's modulus value entered in both, same load definition. Walking the rows, most are marked holds; one is not. At this load the tip has moved far enough that the small-deflection premise of the closed form no longer applies — the discrepancy is geometric nonlinearity, exactly where Euler–Bernoulli is out of range. No theory is wrong; one clearly-named assumption has left its validity window, and that is the whole finding.
How it works¶
- List premises from both sides. Enumerate the assumptions the old formulation needs, the assumptions the new one makes, and — as their own rows — the measurement and translation premises that make the two commensurable at all.
- Mark each against the failing regime. For every row: holds / violated / unknown, with the evidence for the mark. Rows marked unknown become follow-up work, not silent passes.
- Find the minimal breaking set. Identify the smallest set of violated premises that could account for the observed divergence, rather than blaming the first one spotted.
- Hand off. The named broken premise feeds boundary extraction and refinement; the worksheet stops at the diagnosis.
Tuning parameters¶
- Assumption granularity — how finely premises are split. Finer rows localize the break precisely but lengthen the audit; coarse rows are quick but can bury the real culprit inside a compound premise.
- Commensurability scope — how many measurement/translation rows to carry. Thorough translation rows catch the frequent unit/schema false alarm; too many turn the audit into clerical work.
- Evidence bar for "holds" — how much proof a holds mark requires. A high bar prevents wishful ticking; a low bar makes the worksheet fast but hollow.
- Shared-vs-divergent tagging — whether each row is a premise both formulations share or one only the new one makes, which sharpens where the two genuinely part ways.
When it helps, and when it misleads¶
Its strength is that it intercepts the archetype's number-one failure — false violation from translation error — before anyone rewrites a theory, by forcing the boring premises (units, denominators, definitions) into the same disciplined pass as the interesting ones. It turns "the models disagree" into "premise 4 is out of range here," which is a repairable statement.
Its failure mode is that no checklist can enumerate every auxiliary premise a comparison rests on. This is the practical face of the Duhem–Quine problem[n1]: a prediction is never tested in isolation but together with a web of background assumptions, so a violation can always be blamed on an unlisted one — and a worksheet can lull a team into believing the list is complete. The classic misuse is checklist theater: ticking every row holds without evidence, then declaring the theory sound. The guarding discipline is to keep a standing open row for "unstated assumptions we haven't surfaced," treat the worksheet as an attempt to falsify comparability rather than to bless it, and re-open it whenever a downstream repair fails.
How it implements the components¶
The worksheet fills the premise-checking part of the machinery:
assumption_ledger— it is the ledger: the enumerated premises that must hold for correspondence, each with a status in the failing regime.measurement_translation_bridge— the commensurability rows (units, denominators, schema, operational definitions) are the concrete bridge check, confirming the two formulations are spoken in the same language before any divergence is believed.
It does not implement boundary_condition_extractor — turning the named broken premise into a refined statement of where the formulation may be trusted is the job of Boundary Condition Memo. The worksheet finds *which assumption broke; the memo states the resulting where.*
Related¶
- Instantiates: Correspondence Violation Detection and Theory Refinement — it supplies the premise-level diagnosis the refinement loop needs before it can repair anything.
- Consumes: Anomaly Triage Board — it takes a triaged violation and audits its premises.
- Sibling mechanisms: Anomaly Triage Board · Boundary Condition Memo · Expert Adjudication Panel · Failure Mode Annotation Card · Limiting-Case Test Suite · Overlap-Regime Benchmark Table · Parameter Sweep Matrix · Regression Correspondence Harness · Residual Divergence Map
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: Enumerates, one row per premise, which assumptions must hold for two formulations to correspond and marks which of them fails in the regime where agreement broke, making its operative form a bounded evaluation of existing evidence or work that produces a finding or disposition.
Independent corroboration: The frozen evidence defines Assumption Audit Worksheet as 'Enumerates, one row per premise, which assumptions must hold for two formulations to correspond and marks which of them fails in the regime where agreement broke', so its operative form is Assessment, Review & Assurance.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Philosophy
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Locating correspondence failure in the premises that connect two formulations follows philosophy of science and the Duhem–Quine problem.
Related originating lineages:
- Engineering & Design — Model verification checks boundary conditions, units, and validity windows.
- Physics — The worked use of limiting theories and approximations is characteristic of physical modeling.
- Statistics & Experimental Design — Measurement translation and uncertainty distinguish error from theory failure.
Review resolution: Both reviewers agree that correspondence and auxiliary-premise analysis is philosophical, with engineering, physics, and statistical comparison as formative traditions. The worksheet travels broadly across theory-comparison settings, but multi-domain reach is better calibrated than universal presence.
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 Duhem–Quine thesis (Pierre Duhem, W. V. O. Quine) holds that a hypothesis is never tested alone but only in conjunction with a bundle of auxiliary assumptions, so an experimental failure never uniquely fingers the hypothesis — one can always locate the fault in an auxiliary premise instead. For an audit worksheet this is the reminder that the enumerated list is necessarily incomplete. ↩