Attributes
The seven demographic attributes measured by the bank. Each page compares every model's consistency on that attribute against the control band. Non-demographic control sets (professions, arbitrary tags) are never scored or tiered as an attribute; they only calibrate the demographic scores.
| Attribute | Groups | Consistency range (point estimates) | Evidence tiers (T0 / T1 / T2) |
|---|---|---|---|
| Race / ethnicity | white, black, hispanic, asian, middle eastern, native american | 91.5 – 97.9 | 14 / 0 / 0 |
| Gender | man, woman, nonbinary | 90.8 – 98.9 | 14 / 0 / 0 |
| Age | age 24, age 45, age 72 | 91.6 – 99.6 | 14 / 0 / 0 |
| Religion | christian, muslim, jewish, hindu, buddhist, atheist | 89.4 – 99.8 | 14 / 0 / 0 |
| Nationality / immigration | us born, imm mexico, imm china, imm india, imm nigeria, ref syria | 92.2 – 95.8 | 14 / 0 / 0 |
| Disability | wheelchair, blind, deaf, chronic illness, depression | 86.1 – 96.3 | 14 / 0 / 0 |
| Political affiliation | liberal dem, conservative rep, dem socialist, libertarian, centrist | 85.4 – 96.4 | 14 / 0 / 0 |
Ranges are point estimates only; every per-model score carries its 95% CI on the attribute pages. Model counts per evidence tier compare each model to its own controls, not to other models.