Skip to content

Category Audit

An audit — instantiates Social Reality Construction Audit

Pins down a single category — how it is defined, who assigns it, how it is recorded, and what it triggers — so its taken-for-granted boundaries become inspectable.

A Category Audit fixes on one category as its object and inspects the category's internal mechanics: the working definition (often several), the assignment procedure — who decides, on what evidence, with how much discretion — where it gets recorded, and what actions the label triggers. It answers "what actually is this category, and how is it applied in practice," which is a different question from "what does it do to people downstream" or "is it fair." Its defining move is to name the constructed object precisely and walk its assignment routine, turning a word everyone uses into something you can inspect.

Example

A university treats "first-generation student" as a settled category. A Category Audit finds it is nothing of the kind: admissions applies "neither parent holds a four-year degree," a scholarship office uses "neither parent attended college at all," and advising relies on self-report. Assignment is a single unverified checkbox on the application, sometimes back-filled by staff. It is recorded in three systems that never reconcile, and it triggers a summer-bridge invitation in one and a scholarship in another. The audit's output: what looked like one category is three different categories wearing one name, assigned by inconsistent routines. Classification, as Bowker and Star put it, is infrastructure — quiet, ubiquitous, and consequential.[1]

How it works

The distinguishing discipline is to treat the category as the unit of analysis and trace its operational life:

  • Fix the object and collect every working definition — the official one and every de-facto one in use.
  • Map the assignment routine — who applies the label, on what evidence, with what discretion.
  • Find where it is recorded — and whether the records agree.
  • List the actions the label triggers — what happens once someone is in the category.

It stops at describing the category's mechanics; it does not score what those mechanics do to people.

Tuning parameters

  • Definitional resolution — accept the official definition versus collect every definition actually in use. Higher resolution exposes drift but takes more digging.
  • Assignment focus — audit the written rule versus shadow the real assignment behavior. Watching real assignment catches the discretion the rule hides.
  • System coverage — one record system versus every place the category lives. Broader coverage finds the reconciliation gaps.
  • Edge-case sampling — how many boundary cases to pull. Edge cases reveal where the definition actually breaks.

When it helps, and when it misleads

Its strength is turning a word everyone uses into an inspectable object, and it routinely reveals that a "single" category is several inconsistent ones, or that assignment rests on unexamined discretion. Its failure mode is auditing the definition on paper while ignoring how the label is really assigned, and its classic misuse is stopping at "the category is messy" as if messiness were harmless — the consequences are a separate job, but the audit should not pretend they don't exist. The discipline that guards against it is to audit assignment-in-practice, not just the definition, and to hand the effects question explicitly to the impact review.

How it implements the components

Category Audit realizes the object-definition side of the archetype — it fixes and inspects the category, but it does not follow its consequences:

  • constructed_category_or_practice — names and pins the object; the audit's precondition and its primary deliverable.
  • interaction_routine_map — maps the assignment, recording, and action routines through which the category is enacted.

It does not assess the category's downstream effects on access and standing (boundary_and_status_effect_map, affected_party_meaning_map → Classification Impact Review) or trace where the category came from (historical_genealogy → Practice Genealogy).

Notes

It is the upstream partner of Classification Impact Review, which takes the audited category and follows what it does. This one only inspects the category; it neither judges nor changes it.

References

[1] Geoffrey Bowker and Susan Leigh Star, Sorting Things Out: Classification and Its Consequences (1999) — show that classification systems are pervasive, largely invisible infrastructure whose definitional choices carry real consequences. Named because a Category Audit makes that infrastructure visible for one category at a time. ↩