United StatesTechnology and telecom2019 – ongoingMeta Platforms
Machine-learning housing-ad targeting and delivery
Meta withdrew Special Ad Audience for housing ads and accepted court oversight of ad delivery
Problem
Meta predicted which users would engage with advertisements. Federal authorities alleged that its housing-ad algorithms discriminated through targeting and delivery.
Approach
Meta modified its lookalike targeting tool for housing advertisements in 2019. A subsequent federal settlement required further changes. By January 2023, Meta had stopped housing-ad delivery through Special Ad Audience and built a Variance Reduction System to address differences between eligible and actual audiences.
Results
No published figures
Takeaway
Restrictions on advertiser targeting did not resolve the separate discrimination allegations concerning automated delivery.
Categories
- Size
- Large enterprise
- Industry
- Technology and telecom
- Business function
- Marketing and sales
- Innovation type
- Process innovation
- Adoption
- Built in-house
- Evidence
- Regulator, audit or court
Sources
Source summaries
The department confirmed withdrawal of Special Ad Audience for housing advertisements. It described compliance targets for the replacement system and independent monitoring under court oversight.
The complaint described machine-learning systems used for audience targeting and advertisement delivery. It alleged that changes made to housing-ad targeting in 2019 still allowed discrimination involving protected characteristics.
Full details are in the original sources.