Relevant Asymmetry Test¶
Diagnostic relevance test — instantiates Directed Asymmetry Mapping and Calibration
Asks whether a real difference between the two sides is actually relevant to the treatment in question — the gate that separates a warranted asymmetry from bare prejudice or arbitrary privilege.
Detecting that two sides differ is not the same as being allowed to treat them differently. The Relevant Asymmetry Test is the gate in between: given a proposed asymmetric treatment, it asks whether the difference being leaned on is relevant to that specific treatment — causally connected to the purpose at hand — or merely a real-but-irrelevant difference, or a proxy for something impermissible. Its defining move is the demand for a warrant: the difference must earn the differential treatment by being pertinent to it, not just by existing. A test that passes certifies the asymmetry and records why; a test that fails marks the treatment as arbitrary privilege or bias wearing the costume of a "difference."
Example¶
An auto insurer is setting premiums and weighing two candidate factors that would price drivers differently. Factor one is annual mileage. Run the test: is mileage relevant to the risk being priced? Yes — more miles driven is causally tied to more exposure to collisions, and the connection is direct and explainable. The asymmetry earns a warrant, and the reason is written down: higher mileage → higher expected claims, priced accordingly.
Factor two is a variable that merely correlates with a protected characteristic while having no causal bearing on driving risk. It might "improve the model," but under the test it fails: the difference it captures is not relevant to the thing being priced; it is a proxy. Treating drivers differently on it would be unjustified asymmetry dressed as actuarial rigour. The output is not just yes/no — it is a reasoned record: mileage in, warranted and documented; the proxy out, with the failure logged so the same factor cannot quietly reappear next cycle.
How it works¶
- State the treatment and the difference together. Name the exact differential action and the exact difference offered to justify it; a difference can be relevant to one treatment and irrelevant to another, so they must be tested as a pair.
- Test for pertinence, not mere existence. Ask whether the difference is causally connected to the purpose — would the treatment still make sense if this were the only thing you knew? — rather than whether the two sides differ at all.
- Warrant or reject, with reasons. A pass yields an explicit justification tied to the purpose; a fail names the difference as irrelevant, proxy, or pretext, and both outcomes are recorded so they can be re-examined.
Tuning parameters¶
- Relevance standard — how tight the required connection is (loosely correlated, statistically predictive, or causally necessary). Stricter standards reject more proxies but may forbid genuinely useful distinctions.
- Burden of proof — which way the default runs: asymmetry allowed unless shown irrelevant, or forbidden unless shown relevant. The second is the stance to take where the weaker side bears the cost of a wrong call.
- Purpose specification — how narrowly the "treatment's purpose" is defined, since a vague purpose lets almost any difference look relevant.
- Proxy scrutiny — how hard you probe whether an apparently relevant factor is standing in for an impermissible one.
When it helps, and when it misleads¶
Its strength is that it holds the line the archetype cares most about: it lets warranted asymmetries through (a real, pertinent difference) while catching unjustified ones (privilege, stigma, or proxy discrimination sold as a "difference"). It is the operational form of the old principle to treat like cases alike and unlike cases differently in proportion to their relevant difference — with the weight resting on relevant.[n1]
Its failure modes cluster around the word "relevant." The test is routinely run backwards — a decision already made, then a relevant-sounding difference reverse-engineered to warrant it (the pretextual warrant). A loosely drawn purpose can make almost any difference pass, and a proxy can masquerade as directly relevant. Note too what the test does not do: it certifies relevance, not proportion — a genuinely relevant difference can still license a wildly excessive response. The discipline is to fix the purpose and the standard before examining the difference, demand the causal story out loud, and treat a passed test as a licence to differ, not a licence to differ by any amount.
How it implements the components¶
The Relevant Asymmetry Test fills the archetype's justification slots — what a judgement, not a map or a monitor, can fill:
relevant_difference_warrant— its core output: the reasoned certificate that a difference is (or is not) pertinent to the treatment proposed.justified_asymmetry_record— a pass writes the case's justification down as a durable record; the standing collection of such records across the system is maintained by the Asymmetry Exception Register, which this test feeds.
It does not detect the asymmetry in the first place (side_swap_counterfactual → Side-Swap Test; oriented_relation_map → Directed Relation Matrix), scan the whole system for live false-symmetry errors (false_symmetry_guard → False Symmetry Review), or turn a warrant into per-role rules (role_specific_obligation_map → Role-Specific Policy Table).
Related¶
- Instantiates: Directed Asymmetry Mapping and Calibration — it is the gate that turns a detected asymmetry into a warranted or rejected one.
- Consumes: Side-Swap Test and Directed Relation Matrix surface the candidate asymmetry it then adjudicates.
- Sibling mechanisms: Side-Swap Test · Role-Specific Policy Table · Asymmetry Exception Register · False Symmetry Review · Directed Relation Matrix
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: Relevant Asymmetry Test operates as a bounded evaluation of existing evidence or work that produces a finding or disposition because it asks whether a real difference between the two sides is actually relevant to the treatment in question — the gate that separates a warranted asymmetry from bare prejudice or arbitrary privilege.
Independent corroboration: The frozen evidence defines Relevant Asymmetry Test as 'Asks whether a real difference between the two sides is actually relevant to the treatment in question — the gate that separates a warranted asymmetry from bare prejudice or arbitrary privilege', so its operative form is Assessment, Review & Assurance.
Nearest alternative: Decision, Gate & Allocation — Relevant Asymmetry Test includes features of a case-specific gate, selection, routing, prioritization, or resource disposition, but its defining operation is a bounded evaluation of existing evidence or work that produces a finding or disposition.
Review outcome: Independent reviewer agreement; medium confidence.
Origin Attribution¶
Primary origin: Philosophy
Origin pattern: Convergent development
Present-day reach: Universal
Rationale: The formal-equality problem asks which differences are normatively relevant enough to justify unequal treatment; law and policy instantiate that prior ethical test.
Related originating lineages:
- Law & Governance — law_governance contributes rights, duties, authorization, disclosure, and accountable procedure to the mechanism’s formative or independently convergent form; that contribution does not displace the primary philosophy lineage.
- Public Administration & Policy — public_administration_policy contributes program oversight, implementation, and public accountability to the mechanism’s formative or independently convergent form; that contribution does not displace the primary philosophy lineage.
Review resolution: The blind reviewers disagreed on primary lineage; authoritative research supports philosophy over the competing primary. The formal-equality problem asks which differences are normatively relevant enough to justify unequal treatment; law and policy instantiate that prior ethical test. The cited Stanford Encyclopedia of Philosophy: Equality provides direct evidence for that defining form. Alternates are retained only where they contributed an independent formative tradition, while domain_reach=universal records later transfer separately from historical origin.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
Notes¶
Relevance is necessary but not sufficient. A difference can pass this test — genuinely pertinent to the purpose — and the resulting treatment can still be disproportionate, poorly bounded, or in need of an offset. This test hands a licence to differ downstream; how far to differ, and what to give the disadvantaged side in return, belongs to the Role-Specific Policy Table and the compensating controls it encodes.
[n1] The formal principle of equality — treat like cases alike, and unlike cases differently in proportion to their relevant differences — is a classical maxim of justice associated with Aristotle. It is purely formal: it fixes the shape of fairness but says nothing about which differences count, which is exactly the judgement this test supplies. ↩