Facebook has quietly removed “multicultural affinity” ad targeting option that allowed targeting by race, after years of charges of illegal discrimination
meaning moderators typically couldn't see a user's race, but its advertising algo could. https://www.getrevue.co/... @themarkup : This month, Facebook finally removed the “multicultural affinity” targeting category after we discovered the company was again allowing racial profiling in a job ad on its platform. More on what @jeremybmerrill found, and Facebook's response: https://www.getrevue.co/... Julia Angwin / @juliaangwin : I went on vacation and came back to find that Facebook had quietly ended racial ad profiling. It's a big change. Over the years, Facebook's racial ad targeting has been repeatedly found to enable illegal discrimination. https://www.getrevue.co/...
Context & Ripple Effects
This closes a four-year loop that began when ProPublica flagged the abuse potential of Facebook's "ethnic affinity" targeting in 2016. Facebook's first response was narrow: a pledge to bar the category only in housing, employment, and credit ads (announced November 2016), followed by a temporary halt on letting advertisers exclude racial or ethnic groups while it studied the practice.
The gap persisted — a year later, housing ads excluding protected demographics were still being approved. What changed this month is scope: rather than patching verticals or exclusion settings, Facebook deleted the multicultural affinity category outright, after The Markup found a job ad again using racial profiling. Removing the category ends both inclusion- and exclusion-based racial targeting at once.
First-order effects
- Advertisers who built audiences on multicultural affinity segments lose that lever immediately and must rebuild targeting around non-demographic signals like interests and behavior.
- Facebook's self-service ad tools no longer expose an inferred-race field, closing the discrepancy where moderators couldn't see a user's race but the ad system could.
Second-order effects
- Advertisers seeking demographic reach will migrate to proxy signals — geography, interest clusters, lookalikes — shifting the discrimination question from explicit categories to harder-to-audit inference methods.
- Civil-rights scrutiny now pivots to Facebook's remaining targeting and exclusion controls, since the 2016–2017 record shows vertical-specific bans alone did not stop discriminatory housing and employment ads.
Third-order effects
- The pattern — investigative finding, partial fix, recurrence, then structural removal — points toward ad platforms treating inferred sensitive attributes as liabilities they retire wholesale rather than police per-vertical, likely under growing regulatory pressure.
- If proxy-based targeting becomes the norm, enforcement will need to audit ad delivery outcomes rather than targeting inputs, reshaping how discrimination claims against platforms are even testable.
The trend: Social ad platforms are being pushed, by repeated investigative exposure, from policing discriminatory uses of demographic targeting to eliminating inferred-race categories altogether.