Value Audit¶
Method — instantiates Normative Assumption Explicitness
Reviews a policy, model, metric, or process to identify hidden value priorities, displaced alternatives, affected parties, and unsupported legitimacy claims.
A Value Audit starts from a decision that has already been made and works backward, treating the finished artifact — the policy, model, ranking, or process — as evidence to be interrogated rather than trusted. Its whole job is to reconstruct the ought-claim that was smuggled in as if it were a technical necessity: what the design actually rewards, which credible alternatives it beat without ever arguing against them, whom it lands on, and whether its claim to be "objective" or "data-driven" survives inspection. Where a values statement declares and a decision record stores, the audit discovers. It is adversarial by construction, in the way a financial audit is: it assumes the neutrality claim is a hypothesis to be tested, not a fact to be accepted.
Example¶
A company commissions a Value Audit of its last two promotion cycles, which leadership has described as strictly meritocratic. The auditor does not ask what the values should be; she reconstructs what "merit" operationally meant by looking at who actually advanced. The pattern is unmistakable: promotions tracked visibility and logged hours far more than delivered outcomes. That is the hidden normative claim — merit ought to mean being seen working — and nobody had ever stated it. She then lists the credible standards it quietly displaced: demonstrated potential, mentoring load carried, equity of access to high-visibility projects. She maps who bore the cost — caregivers and quieter contributors, disproportionately. Finally she stress-tests the legitimacy claim itself: the process called itself "data-driven," but the data was peer nominations, which encode the very visibility bias in question.
The audit's deliverable is not a verdict but a findings memo: merit, as your system operationalizes it, rewards self-promotion, and the "objective, data-driven" label is unsupported by the data you used. That sentence turns a defended process into a contestable one.
How it works¶
- Fix the decision surface. Name exactly what is under audit — one metric, one model, one policy — so the sweep does not diffuse into everything.
- Reconstruct the ought-claim from behavior. Infer the operative value from what the artifact actually rewards, not from what its authors say it values.
- Enumerate the displaced alternatives. List the credible standards the decision beat silently, so the choice reads as a choice.
- Test the legitimacy claim. Take each "neutral / objective / evidence-based" assertion as a hypothesis and check it against the actual evidence and method.
- Trace who it lands on. Identify the parties who benefit and who bear the cost of the operative value.
Tuning parameters¶
- Auditor independence — an outside reviewer sees more but knows the context less; an insider knows more but is easier to capture.
- Surface scope — one metric versus a whole program. Wider scope catches interactions but dilutes depth.
- Adversarial stance — how hard the neutrality claims are pushed. Harder pressure surfaces more but strains cooperation.
- Evidence standard — how much proof a "hidden priority" finding requires before it is reported.
When it helps, and when it misleads¶
Its strength is catching false neutrality after the fact, in exactly the place nobody thought to look, because it reads outcomes rather than intentions. It is the retrospective mirror of value-sensitive design, which tries to account for human values while a system is being built.[n1]
Its failure mode is that findings can be produced and then filed unread, or curdle into after-the-fact blame that teaches everyone to hide their reasoning next time. The classic misuse is audit-washing — commissioning a review whose real purpose is to launder a decision already defended, with terms of reference narrow enough that nothing can be found. The guarding discipline is to pre-commit to acting on findings, keep the auditor genuinely independent, and route each finding into a durable record rather than a slide that disappears.
How it implements the components¶
normative_claim— reconstructs the operative ought-claim from what the artifact rewards, naming the value that was passing as technical necessity.competing_norm— enumerates the credible alternative standards the decision displaced without argument.legitimacy_check— treats each neutrality or "objectivity" claim as a hypothesis and tests it against the actual method and evidence.affected_party_map— traces who benefits and who bears the cost of the operative value.
It surfaces value premises but neither preserves nor convenes: the durable assumption_record and review_trigger belong to Decision Record, and the live contestation_channel belongs to Stakeholder Deliberation. Where the surface under audit is a single metric, the narrower fact_value_boundary dissection is Metric Value Review's — the Audit sweeps a whole decision surface, the Review stays inside one number.
Related¶
- Instantiates: Normative Assumption Explicitness — supplies the surfaced ought-claims, displaced alternatives, and tested legitimacy claims the rest of the archetype governs.
- Sibling mechanisms: Metric Value Review · Decision Record · Policy Rationale Statement · Stakeholder Deliberation · Ethics Checklist · Model Card Value Section · Values Statement · Ethical Impact Assessment
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: Value Audit operates as a bounded evaluation of existing evidence or work that produces a finding or disposition because it reviews a policy, model, metric, or process to identify hidden value priorities, displaced alternatives, affected parties, and unsupported legitimacy claims.
Independent corroboration: The frozen evidence defines Value Audit as 'Reviews a policy, model, metric, or process to identify hidden value priorities, displaced alternatives, affected parties, and unsupported legitimacy claims', so its operative form is Assessment, Review & Assurance.
Nearest alternative: Analysis, Modeling & Optimization — Value Audit includes features of an analytical, modeling, inference, comparison, or optimization procedure that derives insight or a solution, 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: Historically ambiguous
Present-day reach: Universal
Rationale: Stanford Encyclopedia of Philosophy, Value Theory documents that philosophy distinguishes kinds of value and the reasons by which evaluative judgments are warranted. This is direct, mechanism-specific evidence for philosophy as the best-evidenced historical home of the operation—Reviews a policy, model, metric, or process to identify hidden value priorities, displaced alternatives, affected parties, and unsupported legitimacy claims.—rather than evidence merely that the operation is useful there. The retained alternates record genuine adjacent lineages; later portability is represented separately by domain_reach=universal.
Related originating lineages:
- Accounting & Auditing — Accounting and audit's variance, evidence, ledger, and assurance tradition contributes a separate formative lineage to the mechanism's value audit logic.
- Economics & Finance — Economics Finance supplies a historically relevant adjacent lineage or formative practice for the operation—Reviews a policy, model, metric, or process to identify hidden value priorities, displaced alternatives, affected parties, and unsupported legitimacy claims.—but the adjudicated evidence more directly locates the defining lineage in philosophy.
- Mathematics — Mathematical modeling, proof, and abstract-structure practice supplies a parallel or contributing lineage for the mechanism's defining operation: reviews a policy, model, metric, or process to identify hidden value priorities, displaced alternatives, affected parties, and unsupported legitimacy claims.
- Public Administration & Policy — Public administration, policy implementation, and program oversight supplies a parallel or contributing lineage for the mechanism's defining operation: reviews a policy, model, metric, or process to identify hidden value priorities, displaced alternatives, affected parties, and unsupported legitimacy claims.
- Ethics of Technology & AI Governance — Technology ethics and ai governance supplies a parallel or contributing lineage for the mechanism's defining operation: reviews a policy, model, metric, or process to identify hidden value priorities, displaced alternatives, affected parties, and unsupported legitimacy claims.
Review resolution: The blind reviewers disagree on primary lineage (economics_finance versus philosophy). The defining operation is: Reviews a policy, model, metric, or process to identify hidden value priorities, displaced alternatives, affected parties, and unsupported legitimacy claims. The researched Stanford Encyclopedia of Philosophy, Value Theory establishes that philosophy distinguishes kinds of value and the reasons by which evaluative judgments are warranted. That source therefore supports philosophy as the historical origin. economics finance remains in the uncapped alternates where it contributes a formative practice, but application or governance is not itself proof of origin. origin_mode=historically_ambiguous records lineage construction; domain_reach=universal separately records later applicability.
Attribution caveat: Value audit spans philosophy, public policy, and technology ethics.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; medium confidence.
Sources consulted:
Notes¶
[n1] Value-sensitive design — a design methodology (associated with Batya Friedman) that treats human values as a first-class concern throughout the building of a system. A Value Audit is its retrospective counterpart: it recovers the values a finished system already embodies rather than shaping them up front. ↩