Category Impact Assessment¶
Impact assessment — instantiates Implicit Bias in Knowledge Structure
Traces, for each category choice, how it changes concrete downstream decisions — eligibility, ranking, routing, funding — and who holds the authority those choices feed.
A Category Impact Assessment answers one question about a knowledge structure: what decisions change because it cuts the world this way, and who wields the power those decisions carry? Its defining orientation is forward and consequential — it starts from a category and follows it outward through the machinery it feeds, rather than inward into its labels or assumptions. Two categories can look equally reasonable on the page and have wildly different footprints once you trace where they route people, money, and attention; this mechanism surfaces that footprint before a structure ships or changes. It pairs the consequence trace with a map of whose authority each category feeds — because a category that quietly hands discretion to one office is a different object from an identical-looking one that does not — and it leaves behind indicators to watch the impact over time.
Example¶
A university is revising the "dependency status" categories that decide how financial aid is calculated — whether a student is assessed on their own income or their parents'. The categories look administrative. A Category Impact Assessment traces them forward. A student estranged from her parents but not legally emancipated is classed "dependent," so aid is computed on parental income she cannot access — routing her to a denial she has no standard way to appeal. A student supporting younger siblings is "independent" for income but the category ignores his dependents, understating his need. The assessment does not ask whether these labels sound fair; it follows each to the dollar figure and the eligibility gate it produces.
Then it maps the authority: the "dependency override" that could fix the estranged student's case sits entirely with a single financial-aid officer's discretion, with no published criteria — a concentration of power the category structure creates and hides. The assessment's deliverable is a decision-footprint table (category → aid computation → eligibility outcome → who decides) plus monitoring indicators: track override request and grant rates, and the denial rate for students flagged as potentially estranged. It stops at the footprint; it does not itself rewrite the categories or convene anyone — it shows what the current cut does and who controls the exceptions.
How it works¶
The assessment is defined by tracing outward and mapping the power at each hop:
- Enumerate the decisions each category feeds. For every category, list the concrete downstream actions it drives — eligibility, ranking, routing, funding, escalation — not its meaning.
- Trace representative cases to outcomes. Follow real or realistic cases through the chain to the decision they land in, quantifying the effect where possible (the dollar, the wait, the denial).
- Map the authority at each decision. Record who holds discretion at each hop, and flag where a category concentrates unreviewable power or hides an exception pathway.
- Set impact indicators. Leave behind measurable signals — override rates, denial rates by category, routing skew — so the footprint can be watched rather than assumed stable.
Tuning parameters¶
- Trace depth — how many decision hops downstream to follow. One hop is cheap and usually enough; deeper traces catch laundered effects but risk over-attributing outcomes to the category.
- Quantification effort — from a qualitative footprint to a modeled estimate of who is affected and by how much. Numbers persuade and enable monitoring; they also invite false precision on effects that resist measurement.
- Authority resolution — how finely to map discretion (a whole office versus one named role). Finer resolution exposes hidden single points of power but is intrusive and quickly dated.
- Indicator burden — how many impact signals to stand up. More indicators catch more drift but cost data plumbing and can bury the one metric that matters.
When it helps, and when it misleads¶
Its strength is separating structural bias from mere vocabulary preference: a category only matters to the degree it changes a decision, and this mechanism is what proves — or disproves — that a disputed label actually moves outcomes. It is also the mechanism most likely to expose a disparate impact, where a facially neutral category produces systematically unequal effects downstream.[n1] And by mapping authority, it catches the quiet governance failure of a category that hands one office unreviewable discretion.
Its failure mode is over-attribution — blaming a category for an outcome that many factors produced, so that a long enough trace can indict almost any structure. A related misuse is running the assessment as an after-the-fact justification, tracing only the cases that flatter a decision already made. The guarding discipline is to bound the trace to defensible hops, to state the counterfactual (would the outcome differ if the category were cut differently?), and to keep the impact indicators live so the claimed footprint is checked against reality rather than frozen as an argument.
How it implements the components¶
The assessment fills the consequence-and-authority components:
downstream_consequence_trace— its core: following each category outward to the concrete decisions and outcomes it drives, with the effect quantified where possible.power_or_authority_map— it records who holds discretion at each decision hop and flags where a category concentrates or conceals unreviewable authority.validation_and_monitoring_indicator— it leaves measurable impact signals (override, denial, routing-skew rates) so the footprint is watched over time.
It does not inspect the categories' internal structure or test misfit cases (category_audit, excluded_case_review) — that is taxonomy_bias_audit.md — and it does not gather the affected people's own accounts (stakeholder_perspective_set), which is inclusive_classification_review.md.
Related¶
- Instantiates: Implicit Bias in Knowledge Structure — the assessment is the archetype's consequence-and-power footprint of a category structure.
- Consumes: Taxonomy Bias Audit can supply the flagged categories whose downstream footprint this assessment quantifies.
- Sibling mechanisms: Bias Review Checklist · Taxonomy Bias Audit · Boundary Critique Session · Category Revision Log · Excluded Case Sampling · Inclusive Classification Review · Ontology Critique Workshop · Red-Team Schema Review · Stakeholder Category Review
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: Traces, for each category choice, how it changes concrete downstream decisions — eligibility, ranking, routing, funding — and who holds the authority those choices feed, making its operative form a bounded evaluation of existing evidence or work that produces a finding or disposition.
Independent corroboration: The frozen evidence defines Category Impact Assessment as 'Traces, for each category choice, how it changes concrete downstream decisions — eligibility, ranking, routing, funding — and who holds the authority those choices feed', so its operative form is Assessment, Review & Assurance.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Law & Governance
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Disparate-impact and equality law made downstream effects of facially neutral classifications a reviewable object tied to concrete eligibility and allocation decisions.
Related originating lineages:
- Public Administration & Policy — Impact assessment practice traces decisions, affected groups, authority, and mitigation through administrative systems.
- Sociology & Anthropology — Classification sociology explains how labels produce status, opportunity, and institutional consequences.
Review resolution: Law is primary because disparate-impact analysis made the downstream effects of facially neutral classifications reviewable. Classification sociology explains institutional consequences, while public-administration impact assessment supplies decision-chain, authority, and monitoring practice; the source intentionally synthesizes all three.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
[n1] Disparate impact is the effect of a facially neutral rule or category that nonetheless produces systematically unequal outcomes for a protected group. It names precisely what a consequence trace is built to detect: bias that lives not in a category's wording but in what the category does to different people downstream. ↩