Facebook isn’t supposed to let advertisers target by things like age or gender for employment ads anymore, but a new study showed that the social network’s ad algorithm is doing it anyway.
More than two years after Facebook disabled that particular feature for housing and employment listings, researchers at the University of Southern California found that certain job listings mysteriously still show a noticeable skew in who sees them. The study’s findings would be illegal under federal employment discrimination law.
“Our system takes into account many signals to try and serve people ads they will be most interested in, but we understand the concerns raised in the report,” a Facebook spokesperson said in a statement responding to the study, shared with Mashable. “We’ve taken meaningful steps to address issues of discrimination in ads and have teams working on ads fairness today. We’re continuing to work closely with the civil rights community, regulators, and academics on these important matters.”
To understand the study, one must first understand an interesting carve-out in Title VII, the U.S. law that prohibits employment discrimination on the basis of race, gender, and so on. Employers are still technically allowed to hire based on “bona fide occupational qualifications.” That means a company that makes men’s clothing is legally allowed to only consider men for modeling positions, when that would be discrimination in most other fields.
While previous studies that reached the same conclusion about Facebook’s ad algorithms used different methodology, the USC study focused on this by creating job listings for jobs that exhibit real-life gender demographic skews despite no real difference in qualifications. For example, Domino’s Pizza delivery drivers are overwhelmingly male while Instacart grocery deliverers are predominantly female, per the study.
When the researchers created job listings for Domino’s drivers and Instacart shoppers and ran them at the same time to the same audiences, the former were largely delivered to men and the latter to women by Facebook’s backend advertising technology.
This held true even with listings for software engineer and retail sales associate jobs that also carry the same real-life gender skew. Even when choosing to optimize the ad campaign for the most possible views (as opposed to the most likely click-throughs, another option Facebook offers advertisers), the gender gap still persisted. Interestingly, none of this happened when the same methodology was applied to LinkedIn’s listing service.
We’ve known for years that Facebook’s job ads technology allows for direct or indirect discrimination like this, but the fact that it still happens like this without human intervention years after Facebook promised to do better is concerning. Finding good work is difficult enough without some faceless algorithm keeping job listings away from you on the basis of your identity.