We need gender data to test for fairness.
Gender data can be patchy. So, when we don’t have it, it’s tempting to use titles (a.k.a. prefixes). But there are problems with doing this, including:
Dr, Prof, Rev, Capt don't tell us anything about gender. That would be manageable if those records were spread evenly, but they aren't. Professional titles cluster in particular occupations and income brackets, so whichever way we treat them, we're skewing the population.
Mr is usually first in the dropdown, so it picks up mis-clicks and defaults. If typed in, it’s easy to miskey Mrs as Mr, or the other way around. Blanks are common. Titles are rarely updated when a customer's circumstances change.
In some systems, the title is auto-populated from the gender field at onboarding. Inferring gender from the title then gives us back what we already had, with the errors added by an extra step.
For simplicity, we’ve used the term gender. This doesn’t mean the same thing everywhere. What's recorded at birth may be called “sex”, and gender may or may not be recorded separately.
Where both sex and gender are used, they may be recorded in the same field, in separate fields, or inconsistently across systems. What each one means depends on where and when the data was collected. If we have only one title in our database, do we know whether that relates to sex or to gender?
All this means that using titles or prefixes to infer gender is risky. Better to avoid it if we can. If we decide to use it, we need to be careful how we do it and what we conclude from any subsequent analysis.
Disclaimer: The info in this article is not legal advice. It may not be relevant to your circumstances. It was written for specific contexts within banks and insurers, may not apply to other contexts, and may not be relevant to other types of organisations.