An honest answer.
YES, IN A SPECIFIC SENSE
It reflects patterns in its training data, including unfair ones.
Not because it holds views, but because it reproduces what it learned.
WHERE IT SHOWS
Assumptions about who does what Which contexts are treated as default Performance differences between groups Stereotypes in generated content
WHY REMOVING A FIELD DOES NOT FIX IT
Other variables correlate. Postcode, employer, school and name all carry information.
A system with no protected fields can still produce discriminatory outcomes.
WHAT THIS MEANS PRACTICALLY
For decisions about people, test outcomes by group, not overall accuracy.
For generated content, review for assumptions you did not intend.
FOR NIGERIAN USERS
Content and defaults frequently assume other contexts. Specify.
WHOSE RESPONSIBILITY
Whoever deploys the system.
WHAT TO ASK ANY VENDOR
How was this tested across different groups of people?
