Inclusive Images Library
·
Why we classify representation, and the standard we use
The Inclusive Images Library tags people-depicting illustrations with a representation facet — including gender, age band, and ethnic group — so customers can find, and we can measure gaps in, genuinely diverse depictions of disability. These tags are never presented as filters that imply a hierarchy or a "default" human, and every value passes a community-review gate before it's used to tag a published asset.
Rather than draft a bespoke ethnic-group vocabulary in-house, we adopted an existing, publicly documented, widely used standard. We compared four candidates before choosing:
| Framework | Category count | Scope | Verdict |
|---|---|---|---|
| UK ONS 2021 Census ethnic group classificationChosen | 19 categories | UK national census, publicly documented, revised via public consultation | Chosen — the right size for a searchable representation facet: specific enough to be meaningful, small enough to browse. |
| US OMB Statistical Policy Directive No. 15 (2024 revision) | 7 categories | US federal statistical standard, revised 2024 (added a Middle Eastern/North African category) | Considered — too coarse for representation search and filtering. |
| Australian Standard Classification of Cultural and Ethnic Groups (ASCCEG) | 278 specific groups (9 broad, 28 narrow) | Australian Bureau of Statistics standard, most recently updated 2025 | Considered — the most locally relevant standard, but far too granular for an illustration-tagging facet. |
| The Diversity Style Guide | Not a classification scheme (700+ terminology entries) | A terminology and respectful-language reference built by journalism and DEI professionals | Not a candidate classification — used as a secondary check on individual term wording, not as the source list. |
The UK ONS classification is the right size for a searchable representation facet: specific enough to be meaningful when a customer is searching for a particular depiction, small enough that the full list is browsable rather than overwhelming. The Australian ASCCEG standard — while the most locally relevant to SMP — runs to 278 specific groups, a level of granularity built for census statistics, not illustration tagging. The US OMB standard's 7 categories are the opposite problem: too coarse to be useful as a search filter.
Grouped exactly as the UK Office for National Statistics groups them in the 2021 Census:
Asian, Asian British or Asian Welsh
- Bangladeshi
- Chinese
- Indian
- Pakistani
- Other Asian
Black, Black British, Black Welsh, Caribbean or African
- African
- Caribbean
- Other Black
Mixed or Multiple ethnic groups
- White and Asian
- White and Black African
- White and Black Caribbean
- Other Mixed or Multiple ethnic groups
White
- English, Welsh, Scottish, Northern Irish or British
- Irish
- Gypsy or Irish Traveller
- Roma
- Other White
Other ethnic group
- Arab
- Any other ethnic group
Read the source classification
The Office for National Statistics' own documentation for the Census 2021 ethnic group variable and its classifications.
This is an external, independently governed standard — we did not invent it, and we don't control its categories. Using it as our representation facet still requires a community-review pass before any category is applied to a published asset, consistent with how every other representation value in the library is handled.
For the full taxonomy this facet sits within — subjects, categories, art styles, and the naming grammar used across the library — see the Cohorts reference for how SMP's six testing cohorts relate to the accessibility categories used throughout the platform.