Normative Assumption Explicitness¶
Essence¶
Normative Assumption Explicitness makes hidden value judgments visible before they harden into policy, metrics, models, rankings, thresholds, or designs. The archetype does not ask teams to pretend values are absent. It asks them to stop treating value choices as if they were neutral facts.
The core move is simple: identify what the decision assumes ought to matter, where that standard comes from, who is affected by it, what competing norms were possible, why the selected standard is acceptable in this context, and when it should be reopened.
Compression statement¶
When a decision, metric, model, policy, or design appears objective while depending on unstated judgments about what ought to matter, Normative Assumption Explicitness identifies the normative claims, names their sources, maps affected parties, compares competing norms, documents rationale, and defines how those assumptions can be contested or revisited.
Canonical formula: apparently_neutral_decision + hidden_ought_claims -> normative_claim_inventory + value_sources + affected_party_map + competing_norms + decision_rationale + contestation_and_review
When to Use This Archetype¶
Use this archetype when a decision claims neutrality while allocating benefit, burden, risk, opportunity, credibility, attention, or protection. It is especially useful when a metric defines success, a model automates prioritization, a policy creates eligibility, or a design choice makes one value easy and another value costly.
It is also useful when people seem to be arguing about facts but are actually disagreeing about standards. For example, a team may debate whether a threshold is “accurate enough,” while the deeper question is whether false positives or false negatives are more acceptable for the people affected.
Structural Problem¶
The structural problem is false neutrality. A system presents a decision as technical, objective, scientific, efficient, or evidence-based, but the decision depends on unstated assumptions about what should count, what should be optimized, what risks are acceptable, and whose outcomes matter.
Hidden normative assumptions create several predictable harms. They prevent affected parties from understanding why a decision was made. They shift value conflict into technical language. They make contestation look irrational because the real ought-claim is never named. They also make later revision difficult because no one knows which value premise the original decision depended on.
Intervention Logic¶
The intervention begins by locating the decision surface: a metric, model, rule, ranking, threshold, policy, budget, product choice, or evaluation frame. The next step is to extract the normative claims embedded in that surface. What is being treated as success? What risk is tolerated? What duty is being prioritized? What kind of fairness is assumed? What is being sacrificed?
The archetype then separates empirical claims from value claims. Evidence can show what is likely, but it does not by itself decide what should matter. Once that boundary is visible, the intervention names value sources, maps affected parties, compares competing norms, records the accepted rationale, and defines contestation and review conditions.
The output is not merely a statement of values. It is a governed assumption record: what value premise guided action, why it was accepted, what alternatives were considered, who is affected, where dissent remains, and when the premise should be reopened.
Key Components¶
Normative Assumption Explicitness converts a decision's hidden value premises into an inspectable record by separating what is being claimed, where the claim comes from, who it affects, and how it can be challenged. The Normative Claim names the ought-statement embedded in the decision — what should outrank what — that would otherwise pass as technical necessity. The Value Source identifies the authority behind that claim, whether law, mission, professional duty, community expectation, or informal norm, so the claim's legitimacy can be inspected. The Fact–Value Boundary does the analytic work of distinguishing empirical estimates from value judgments, so evidence is not used to silently settle questions evidence cannot decide. Together these three components turn an apparently neutral decision into a layered, traceable structure.
The remaining components govern the social and procedural surround. The Affected Party Map prevents value rationales from being constructed only from the position of those with institutional power, and a Competing Norm keeps the chosen standard visible alongside the alternatives it displaced — without that contrast, the rationale is too weak to evaluate. The Decision Rationale records why the chosen priority was accepted in this scope, while a Contestation Channel gives affected parties or reviewers a defined way to push back, preventing explicit values from becoming polished justifications for predetermined choices. The Assumption Record preserves the full memory of the intervention — claim, source, parties, alternatives, rationale, and dissent — so later audit and learning can examine the value premise rather than guess at it. Finally, the Review Trigger defines when the assumption must be reopened, recognizing that legitimacy in one context, stakeholder set, or time period does not silently extend to others.
| Component | Description |
|---|---|
| Normative Claim ↗ | A normative claim is the ought-claim hidden inside the decision. It may say that safety should outrank autonomy, that efficiency should outrank customization, that accuracy should outrank explainability, or that consistency should outrank contextual adjustment. Naming this claim prevents it from disguising itself as technical necessity. |
| Value Source ↗ | A value source explains where the standard comes from. It may come from law, policy, mission, professional duty, community expectation, user preference, safety culture, or informal organizational norms. A value source does not automatically settle the issue, but it makes the claim’s authority and legitimacy inspectable. |
| Affected Party Map ↗ | An affected party map identifies who benefits, who bears risk, who is constrained, and who needs a way to contest the assumption. This component keeps values from being defined only by those with technical or institutional power. |
| Competing Norm ↗ | A competing norm is a credible alternative standard. For example, a school assessment might value test performance, learning growth, inclusion, wellbeing, or long-term capability. The intervention is weak if it only states the winning value. It becomes useful when it also shows what alternatives were displaced. |
| Decision Rationale ↗ | The decision rationale records why a particular normative assumption was accepted in context. It should explain the chosen value priority, the rejected alternatives, the scope of the choice, and the conditions that would justify revision. |
| Fact–Value Boundary ↗ | The fact–value boundary separates empirical claims from value claims. A risk model may estimate likelihood; a policy still decides which risks are acceptable. A performance metric may measure speed; a governance decision still decides whether speed should count as success. |
| Contestation Channel ↗ | A contestation channel gives affected parties or reviewers a way to challenge the assumption. Without this channel, explicit values can become a polished explanation for a predetermined choice. |
| Assumption Record ↗ | The assumption record preserves the claims, sources, affected parties, alternatives, rationale, and review triggers. It is the memory of the intervention and supports future audit, learning, and accountability. |
| Review Trigger ↗ | A review trigger defines when the normative assumption should be reopened. Triggers may include new harms, stakeholder change, legal change, metric drift, deployment in a new context, or evidence that the chosen value priority is producing unacceptable effects. |
Common Mechanisms¶
| Mechanism | Description |
|---|---|
| Value Audit ↗ | A value audit reviews a decision surface to find hidden priorities and displaced alternatives. It implements the archetype by moving implied standards into an explicit record. |
| Ethical Impact Assessment ↗ | An ethical impact assessment can surface harms, duties, rights, and value conflicts. It is not the archetype by itself; it is a structured assessment mechanism that supports explicit normative reasoning. |
| Metric Value Review ↗ | A metric value review asks what a metric rewards, ignores, normalizes, or sacrifices. This mechanism is especially important because metrics often appear objective while quietly defining what should count as success. |
| Policy Rationale Statement ↗ | A policy rationale statement records the values and tradeoffs behind a rule. It implements the archetype when it names affected parties, value sources, competing norms, and review conditions rather than merely defending the rule. |
| Stakeholder Deliberation ↗ | Stakeholder deliberation creates a forum for affected parties and decision makers to compare values. It is a mechanism, not the archetype. It matters because normative assumptions often look different from the position of those who bear consequences. |
| Decision Record ↗ | A decision record makes value premises durable. It stores what was assumed, what alternatives were rejected, what dissent remains, and when the assumption should be revisited. |
| Ethics Checklist ↗ | An ethics checklist can prompt useful questions, but it should not be confused with the archetype. A checklist implements this archetype only when the answers are concrete, sourced, contestable, and tied to the actual decision. |
| Model Card Value Section ↗ | A model card value section documents the value assumptions embedded in a technical system. It can identify intended uses, excluded uses, target choices, evaluation priorities, affected groups, and review triggers. |
| Values Statement ↗ | A values statement is an artifact that can support the archetype, but it is not sufficient. “We value fairness” does not make a decision governable until the specific fairness assumption, tradeoff, rationale, and review path are stated. |
Parameter / Tuning Dimensions¶
Explicitness depth determines how much detail is required. A low-stakes decision may need only a brief note. A high-stakes automated decision may need a full assumption record, stakeholder review, alternatives analysis, and formal review triggers.
Participation scope determines who helps surface and contest assumptions. Internal review may be enough for low-risk work. Public policy, healthcare triage, AI screening, or eligibility systems often require affected-party input.
Rationale formality determines whether the output is a lightweight note, decision record, policy rationale, audit report, or formal documentation. More durable or repeated decisions need more durable rationale.
Norm conflict granularity determines how carefully competing values are compared. Some decisions need a short list of alternatives; others need a matrix that shows how options affect autonomy, safety, equity, efficiency, privacy, dignity, and accountability.
Review cadence determines whether the assumption is reviewed once, periodically, or when triggered by harm reports, drift, new stakeholders, or context change.
Invariants to Preserve¶
The first invariant is the fact–value distinction. Evidence can guide a decision, but evidence does not erase the need to say what should matter.
The second invariant is affected-party visibility. The people who bear risk or burden should not disappear from the value rationale.
The third invariant is competing norm visibility. A chosen value should be visible alongside the credible alternatives it displaces.
The fourth invariant is contestability. Explicit assumptions should be open to challenge through defined channels.
The fifth invariant is scope discipline. A value assumption accepted in one context should not silently govern every context.
Target Outcomes¶
A successful use of the archetype produces more honest decision rationale. Decision makers can explain not only what they chose but what standard made the choice acceptable.
It reduces false neutrality. Technical language no longer hides the underlying value choices.
It improves governance of models, metrics, and policies by making targets, thresholds, rules, and priorities reviewable.
It improves stakeholder trust where disagreement is inevitable, because people can see and contest the assumptions shaping outcomes.
It also improves revision. When a decision later fails, reviewers can examine the value premise rather than guessing what the original choice meant.
Tradeoffs¶
The main tradeoff is speed. Normative explicitness takes time, especially when affected parties must be included. The solution is proportionality: use lightweight records for low-stakes decisions and deeper review for high-stakes or contested decisions.
Another tradeoff is contestability versus closure. The goal is not endless debate. The archetype should define how assumptions can be challenged and when enough review has occurred for action.
A third tradeoff is transparency versus strategic ambiguity. Sometimes ambiguity helps a coalition move forward, but hidden values can later produce mistrust when consequences become visible.
A fourth tradeoff is breadth of participation. Wider participation reveals more values and harms, but it requires representation choices, facilitation, and decision rules.
Failure Modes¶
Ethics theater occurs when the organization fills out a checklist or publishes values without changing actual decision rationale. Mitigation: link every value claim to a concrete decision, affected-party implication, competing norm, and review trigger.
Hidden priority remains hidden occurs when a record lists many values but refuses to say which value wins in conflict. Mitigation: require comparison of competing norms and document the accepted priority for the current context.
False consensus occurs when one value source is treated as universal. Mitigation: record dissent, map affected parties, and provide contestation channels.
Over-philosophizing occurs when the process turns into abstract moral debate. Mitigation: anchor the discussion in a decision surface: a rule, metric, model, threshold, policy, or design choice.
Paralysis from infinite contestation occurs when no closure rule exists. Mitigation: pair this archetype with a regress termination rule or decision closure criterion.
Expert capture occurs when technical or legal experts define the values without affected-party knowledge. Mitigation: combine with epistemic inclusion design, lived experience capture, or stakeholder deliberation.
Neighbor Distinctions¶
Procedural Fairness Design designs fair process: voice, notice, impartiality, reasons, and appeal. Normative Assumption Explicitness names the value premises that guide the decision. They often combine but are not the same.
Objective Weighting Governance governs how objectives are weighted in optimization or scoring systems. Normative Assumption Explicitness is broader and includes nonnumeric decisions, policies, metrics, and designs.
Stakeholder Mapping and Engagement identifies and engages affected parties. This archetype uses stakeholder understanding to expose and govern hidden ought-claims.
Goal Congruence Alignment aligns behavior with accepted goals. Normative Assumption Explicitness asks whether the goals or standards themselves contain hidden values.
Epistemic Inclusion Design protects fairness in knowledge production. Normative Assumption Explicitness protects transparency of value premises. Epistemic inclusion may be necessary when excluded knowers can reveal hidden assumptions.
Purpose Alignment Design aligns means with an accepted purpose. Use Normative Assumption Explicitness when the purpose or the values inside that purpose are hidden or contested.
Variants and Near Names¶
Metric Value Review is a variant for metrics, rankings, KPIs, benchmarks, and scores. It asks what the metric values and what it hides.
Policy Rationale Explicitness is a governance variant for rules and administrative decisions. It states why a policy’s value priority is legitimate and revisable.
Algorithmic Value Assumption Audit is a technical-system variant for AI, scoring, ranking, and automation. It surfaces value choices in labels, targets, thresholds, loss functions, and evaluation priorities.
Objective Weighting Disclosure is a merge-review variant because Objective Weighting Governance may deserve separate treatment in optimization-heavy settings.
Near names include value assumption audit, hidden value audit, normative assumption audit, ought-claim disclosure, and value premise transparency. Ethics checklists and values statements should point here as mechanisms or artifacts, not as separate archetypes.
Cross-Domain Examples¶
In AI hiring, a screening tool may define success using a label that rewards prior corporate experience. The archetype asks the team to state the value assumption, compare it with alternatives such as potential and equity, map affected candidates, and revise the rationale or label.
In public health, a vaccination policy may be described as purely scientific. The evidence matters, but the policy also contains assumptions about liberty, collective protection, trust, and acceptable risk. The archetype separates those layers.
In product design, a platform may optimize engagement and call the choice user-centered. Normative explicitness asks whether engagement should be treated as wellbeing, autonomy, revenue, habit formation, or something else.
In education, a school ranking may reward test scores while ignoring learning growth, inclusion, or wellbeing. The archetype makes the values embedded in ranking criteria visible.
In organizational promotion, a process may claim meritocracy while rewarding visibility, hours, self-promotion, or risk-taking. Normative explicitness reveals what “merit” actually means in the system.
Non-Examples¶
A generic ethics seminar is not this archetype unless it changes how a concrete decision records and governs value assumptions.
A values poster is not this archetype. It becomes relevant only when those values are linked to decisions, tradeoffs, affected parties, and review conditions.
A purely statistical confidence interval is not this archetype. It expresses uncertainty, not what outcome should matter.
A stakeholder list without value analysis is not this archetype. It can support the work, but the intervention requires surfacing and governing ought-claims.