Sensitivity-to-Mapping-Change Review¶
Robustness review — instantiates Preimage Set Characterization
Perturbs the mapping, threshold, or parameters and watches which inputs enter or leave the preimage, exposing how fragile the set is and warning downstream users where it will move.
A Sensitivity-to-Mapping-Change Review treats the mapping itself as the variable. Every other mechanism holds the mapping fixed and characterizes the preimage under it; this one nudges the mapping — moves a threshold, retunes a parameter, swaps a rule version — and watches which inputs cross into or out of the set. Its defining product is not the preimage but its fragility: which members are secure, which are marginal, and how far the mapping can shift before the set is materially different. It exists because a preimage computed against one threshold is quietly a claim about that exact threshold, and downstream users who never see the sensitivity will treat a knife-edge result as if it were bedrock.
Example¶
A lender defines "approved applicants" as everyone whose credit score maps above a cutoff of 680 — the preimage of the output approve. A Sensitivity-to-Mapping-Change Review sweeps the cutoff across a band, say 660 to 700, and records how the approved set moves. It finds the set is stable in its core but churns heavily at the edge: shifting the cutoff by ten points moves several thousand applicants across the line, and a cluster of applicants sits at exactly 679–681, flipping status on a rounding decision.
That fragility map is the review's product. The boundary register names the applicants who are "approved" only because the cutoff is 680 and not 690, and the downstream guardrail tells the servicing team not to treat membership as a durable fact for anyone in the churn band — a marginal approval is a different thing from a secure one. Without the review, a small, defensible retuning of the model would silently rewrite thousands of individual determinations, and no one downstream would know which ones were ever solid.
How it works¶
- Identify the mapping's dials. Enumerate the parameters, thresholds, or rule versions that could plausibly change — the axes along which the mapping is uncertain or adjustable.
- Perturb and recompute. Move each dial across a realistic range and recompute the preimage at each setting, holding the output condition fixed.
- Register the movers. Record which inputs stay in across all settings (secure), which never enter (out), and which flip — the boundary cases whose membership is an artifact of the current mapping.
- Emit a durability guardrail. Translate the fragility into a usage warning: which memberships downstream may rely on and which it must treat as provisional.
Tuning parameters¶
- Perturbation range — how far each dial is swung. A wide range stress-tests robustness hard but includes settings no one would really use; a narrow range is realistic but may miss a nearby cliff.
- Perturbation granularity — how finely the range is stepped. Fine steps locate the exact tipping point at higher cost; coarse steps are cheap but can jump over a threshold where the set lurches.
- One-at-a-time vs. joint — whether dials move singly or in combination. Single moves are interpretable but miss interactions; joint moves catch compound cliffs but explode the number of runs.
- Churn threshold — how much membership movement counts as "fragile" worth flagging. A low threshold flags almost everything; a high one only warns on dramatic swings.
When it helps, and when it misleads¶
Its strength is exposing knife-edge preimages before they are trusted: a one-at-a-time sensitivity analysis[1] over the mapping's parameters shows which conclusions survive a defensible retuning and which evaporate, so a downstream decision can lean on the robust part and hedge the fragile part instead of treating the whole set as equally solid.
Its failure mode is a false sense of security from a perturbation design that misses the real axis of change: sweeping the threshold while the true fragility lies in an untouched parameter, or moving dials one at a time when the cliff only appears when two shift together. The classic misuse is running a token sensitivity check, finding the core stable, and certifying the whole preimage as robust while the marginal band that actually drives the decisions was never separated out. The guarding discipline is to perturb the dials that genuinely could move — not the convenient ones — to test joint moves where interactions are plausible, and to publish the boundary band as loudly as the stable core so downstream users can tell a secure membership from a provisional one.
How it implements the components¶
mapping_under_review— its signature: the mapping (its thresholds, parameters, and versions) is the object being perturbed, not held fixed.boundary_case_register— the log of inputs whose membership flips under perturbation, marked as artifacts of the current mapping rather than durable facts.downstream_use_guardrail— the durability warning that tells consumers which memberships they may rely on and which are provisional.
It does not implement collision_and_multiplicity_check — tabulating the set's internal collisions at a single fixed mapping is the job of Collision Analysis Matrix and Fiber Cardinality Count; this review varies the mapping and watches the whole set move, rather than dissecting the set at one setting.
Related¶
- Instantiates: Preimage Set Characterization — the review supplies the fragility map and durability guardrail that keep a threshold-dependent preimage from being trusted as fixed.
- Consumes: Predicate Satisfaction Filter supplies the membership rule this review re-runs under perturbed thresholds.
- Sibling mechanisms: Inverse Lookup Query · Constraint-Solver Backsolve · Predicate Satisfaction Filter · Output-to-Input Traceback Map · Collision Analysis Matrix · Fiber Cardinality Count · Preimage Table · Witness and Counterexample Set
Editorial Notes¶
Form Classification¶
Form family: Analysis, Modeling & Optimization
Rationale: Sensitivity-to-Mapping-Change Review operates as an analytical, modeling, inference, comparison, or optimization procedure that derives insight or a solution because it perturbs the mapping, threshold, or parameters and watches which inputs enter or leave the preimage, exposing how fragile the set is and warning downstream users where it will move.
Independent corroboration: The frozen evidence defines Sensitivity-to-Mapping-Change Review as 'Perturbs the mapping, threshold, or parameters and watches which inputs enter or leave the preimage, exposing how fragile the set is and warning downstream users where it will move', so its operative form is Analysis, Modeling & Optimization.
Nearest alternative: Assessment, Review & Assurance — Sensitivity-to-Mapping-Change Review includes features of a bounded evaluation of existing evidence or work that produces a finding or disposition, but its defining operation is an analytical, modeling, inference, comparison, or optimization procedure that derives insight or a solution.
Review outcome: Independent reviewer agreement; medium confidence.
Origin Attribution¶
Primary origin: Mathematics
Origin pattern: Convergent development
Present-day reach: Universal
Rationale: Recomputing results after changing a mapping, correspondence, or transformation tests structural rather than merely parametric sensitivity. NIST structural-sensitivity work formalizes how altered model structure affects outputs; data and software disciplines implement the mappings.
Related originating lineages:
- Computer Science & Software Engineering — Schema transformations and classifiers operationalize mappings whose membership changes downstream behavior.
- Data Science & Analytics — Data science, analytics, and operational monitoring supplies a parallel or contributing lineage for the mechanism's defining operation: perturbs the mapping, threshold, or parameters and watches which inputs enter or leave the preimage, exposing how fragile the set is and warning downstream users where it will move.
- Library & Information Science — Reclassification and vocabulary changes move records among categories and require impact review.
- Statistics & Experimental Design — Threshold perturbation estimates classification instability.
Review resolution: The blind reviewers disagree on primary lineage (mathematics versus statistics_experimental_design). Authoritative or primary research supports mathematics as the best historical origin: Recomputing results after changing a mapping, correspondence, or transformation tests structural rather than merely parametric sensitivity. NIST structural-sensitivity work formalizes how altered model structure affects outputs; data and software disciplines implement the mappings. The cited NIST, Structural Sensitivity Analysis directly supports the mechanism's defining operation. All independently supported contributing domains are retained without an arbitrary cap. origin_mode=convergent records the lineage relationship, while domain_reach=universal records later applicability separately from provenance.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
References¶
[1] Saltelli, A., Ratto, M., Andres, T., Campolongo, F., Cariboni, J., Gatelli, D., Saisana, M., & Tarantola, S. Global Sensitivity Analysis: The Primer. John Wiley & Sons (2008). Provides one-at-a-time parameter-variation methods as part of sensitivity analysis. registry ↩