Dominance Violation Scan¶
Diagnostic scan — instantiates Revealed Preference Validation Against Indifference Curves
Flags any single choice where an available option was at least as good on every attribute and strictly better on one — a selection no coherent preference should make.
Before you ask what trade-off a choice reveals, you should ask whether it reveals a coherent preference at all. Dominance Violation Scan answers the crudest, most decisive version of that question, one choice at a time: was there an option, right there in the feasible set, that beat the chosen one on every attribute and lost on none? If so, no trade-off can explain the pick — it is dominated, and any preference model fit to it will be fit to noise, confusion, or a data error. Its defining scope is the single occasion, judged attribute-by-attribute against what was available: it needs no history and no trade-off rates, only the option chosen, the options present, and their attribute bundles.
Example¶
An airline's revenue team scans a quarter of booking records for dominated purchases. For each ticket it compares the fare chosen against every other fare that was live in the same session on the same route and dates, across the attributes that matter — price, cabin, layovers, refundability, bag allowance. Most bookings involve real trade-offs (a cheaper fare with a longer layover) and pass. But the scan surfaces a cluster of tickets where the customer paid for a fare that was strictly worse on every axis than another fare visible on the same screen — higher price, more stops, no bag, non-refundable — a pick no rational trade-off can rationalize.
That flag is not a preference finding; it is a red alert. Investigation shows the dominated fares came from a booking-flow bug that hid the better option below the fold on mobile. The scan's value was to isolate the incoherent choices before they poisoned the trade-off estimates — because feeding a dominated pick into a substitution model would have taught it that customers "prefer" paying more for less.
How it works¶
- Fix the comparison set. For each observed choice, take the feasible options available at that instant.
- Compare attribute-wise. Check whether any available option weakly dominates the chosen one — no worse on all attributes, strictly better on at least one.
- Flag, don't rationalize. Mark every dominated choice as a violation; a coherent preference never selects a dominated option, so there is nothing to fit.
- Route to explanation. Hand each flag onward as an anomaly to be diagnosed (data error, hidden option, confusion), not as a trade-off to be modeled.
Tuning parameters¶
- Dominance strictness — strict (better on all) versus weak (no worse on all, better on one). Weak dominance catches more but flags near-ties that may be noise.
- Attribute completeness — how many attributes enter the comparison. Omit one the chooser cared about and a legitimate choice looks dominated; include irrelevant ones and true violations hide.
- Tolerance band — how large an attribute difference must be to "count," to avoid flagging rounding-level gaps.
- Feasibility confidence — how sure you are the dominating option was truly available and known; a phantom option produces a phantom violation.
- Batching granularity — scan per session, per day, or per market, trading precision against coverage.
When it helps, and when it misleads¶
Its strength is that it is a cheap, unambiguous filter: a genuine dominance violation is a logical impossibility for a coherent chooser, so a flag is high-signal and demands explanation — usually a data problem, a hidden option, or a choice-environment defect — before any modeling proceeds.[n1] Catching these early keeps incoherent choices out of the trade-off estimates.
Its failure mode is the incomplete attribute model: a choice looks dominated only because the analyst left out an attribute the chooser valued (brand trust, seat position, an airline's loyalty status). Declaring such a choice irrational is the classic misuse — the irrationality is in the model, not the chooser. The scan is also blind to any incoherence that isn't attribute-wise dominance; a set of choices can each be individually undominated yet collectively contradictory. The guarding discipline is to treat every flag as a hypothesis about a missing attribute or missing option first, and only then as a real anomaly — and to hand cross-choice contradictions to a mechanism built for them.
How it implements the components¶
Dominance Violation Scan realizes one sharp blade of the archetype's consistency machinery — the single-choice dominance test:
consistency_test_suite— it contributes the dominance check: the per-occasion test that a chosen option is not beaten on every attribute.feasible_option_set— it reads the available alternatives at each choice as the comparison set the pick is judged against.attribute_bundle_model— it compares options attribute-by-attribute using their normalized bundles.
It does not test the choice history for cycles or intransitivity across occasions (observed_choice_record as a relational whole) — that is Revealed Preference Consistency Matrix; this scan judges each choice alone against what was on offer, not choices against each other.
Related¶
- Instantiates: Revealed Preference Validation Against Indifference Curves — it is the first-pass coherence filter that keeps dominated picks out of the inference.
- Consumes: Budget Set Reconstruction supplies the feasible set, and Choice Bundle Normalization the attribute bundles it compares.
- Sibling mechanisms: Budget Set Reconstruction · Choice Bundle Normalization · Marginal Substitution Estimator · Indifference Region Visualization · Revealed Preference Consistency Matrix · Preference Reversal Probe · Choice Architecture Confound Audit · Stated vs Revealed Gap Report · Ethical Preference Inference Review
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: Dominance Violation Scan operates as a bounded evaluation of existing evidence or work that produces a finding or disposition because it flags any single choice where an available option was at least as good on every attribute and strictly better on one — a selection no coherent preference should make.
Independent corroboration: The frozen evidence defines Dominance Violation Scan as 'Flags any single choice where an available option was at least as good on every attribute and strictly better on one — a selection no coherent preference should make', so its operative form is Assessment, Review & Assurance.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Economics & Finance
Origin pattern: Single lineage
Present-day reach: Multi-domain
Rationale: Revealed-preference theory established dominance as a basic rationality condition: choosing an available option worse on every valued attribute requires explanation.
Related originating lineages:
- Behavioral Economics — Behavioral decision research operationalized dominance violations as diagnostic departures from rational choice.
- Operations Research — Decision analysis supplied multi-attribute tables and automated dominance comparisons.
Review resolution: Revealed-preference economics established dominance as a rationality condition; behavioral economics made violations diagnostic and operations research made the scan operational.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
Notes¶
[n1] An option dominates another when it is at least as good on every attribute and strictly better on at least one; choosing a dominated option cannot be rationalized by any monotone preference. This is the most basic rationality check in decision analysis — weaker and more universal than transitivity, which is why it makes a good first scan. ↩