Facebook Ad Platform Could Be Inherently Discriminatory, Researchers Say (theregister.co.uk)

Researchers from Northeastern Unviersity, the University of Southern Carolina, and tech accountability non-profit Upturn have released a paper that says Facebook's ad delivery system itself can steer ads intended to be inclusive toward discrimination without explicit intent. "In a paper titled, 'Discrimination through optimization: How Facebook's ad delivery can lead to skewed outcomes,' co-authors Muhammad Ali, Piotr Sapiezynski, Miranda Bogen, Aleksandra Korolova, Alan Mislove, and Aaron Rieke find that advertiser budgets and ad content affect ad delivery, skewing it along gender and racial lines even when neutral ad targeting settings are used," reports The Register. From the report: The researchers found that Facebook ads tend to be shown to men because women tend to click on ads more often, making them more expensive to reach through Facebook's system. That divide becomes apparent when ad budgets are compared, because the ad budget affects ad distribution. As the paper explains, "the higher the daily budget, the smaller the fraction of men in the audience." Such segregation may be appropriate and desirable for certain types of marketing pitches, but when applied to credit, employment and housing ads, the consequences can be problematic.

Ad content -- text and images -- also has a strong effect on whether ads get shown to men or women, even when the bidding strategy is the same and gender-agnostic targeting is used. In particular, the researchers found images had a surprisingly large effect on ad delivery. Ad URL destination has some effect -- an ad pointing to a bodybuilding site and an ad pointing to a cosmetics site had a baseline delivery distribution of 48 percent men and 40 percent men respectively. The addition of a title and headline doesn't change that much. But once the researchers added an image to the ad, the distribution pattern changed, with the bodybuilding site ad reaching an audience that was 75 percent male and the cosmetics ad reaching an audience that was 90 percent female. According to the researchers, their tests suggest, "Facebook has an automated image classification mechanism in place that is used to steer different ads towards different subsets of the user population." "In terms of credit, employment and housing ads, the problem with this system is that it discriminates where it shouldn't: Five ads for lumber industry jobs were delivered to an audience that was more than 90 percent men and more than 70 percent white; five ads for janitorial work were delivered to an audience that was more than 65 percent women and 75 percent black," the report adds. "Housing ads also showed a racial skew."

The latest findings come after years of criticism of Facebook's ad system. Last month, Facebook announced changes to the platform intended to prevent advertisers from deploying unfair credit, employment and housing ads. One week later, the U.S. Department of Housing and Urban Development sued Facebook for violating the Fair Housing Act.

Comments

The descrimination is not theoretical

[Required Snark]

This is the part everyone is ignoring:

"In terms of credit, employment and housing ads, the problem with this system is that it discriminates where it shouldn't: Five ads for lumber industry jobs were delivered to an audience that was more than 90 percent men and more than 70 percent white; five ads for janitorial work were delivered to an audience that was more than 65 percent women and 75 percent black," the report adds. "Housing ads also showed a racial skew."

That is discriminatory and illegal. The researchers looked at lots of factors, including areas where targeting is acceptable practice. They wanted to know if Facebook was using their targeting algorithm where it should not be used. The answer seems to be that Facebook is breaking the law.

All the whining fools questioning the study are too stupid to read the summery and/or too dumb to understand it: typical Slashdot knuckle dragging idiots who still live in their parent's basement.

Re:The descrimination is not theoretical

[Kjella]

Could you elaborate how a generic targeting algorithm can "discriminate".

A generic ad optimization algorithm will create a bunch of proxy variables that effectively work like the non-discriminating factor, even if it's explicitly denied access to that variable. It's actually a lot harder to avoid than to do, basically treat every time you show an ad like an experiment. If you click the ad it counts for every property about you, if you don't click the ad it counts against every property about you. So you say it's not to discriminate on sex, the algorithm doesn't get your gender. But it'll pretty soon figure out whether people who are fans of Justin Bieber and a million other gender-skewed metrics respond to your ad or not.

It's absolutely feasible to build an algorithm that tries to find relevant candidates while respecting an imposed equality, for example that they draw from separate pools so you must have half male and half female ad impressions. But it gets pretty hard when you want for example a jobs ad that doesn't discriminate on ethnicity, like if the population is 50-30-15-5 percent something you want a 50-30-15-5 percent distribution. It gets really hard if you don't want to discriminate on religion, sexual orientation, political stance or something else where you don't actually have the underlying data but you're pretty sure your proxy variables are effectively filtering on that anyway.

If you're going to reply with "how's an algorithm that's only preserving the status quo or amplifying a feedback loop that's already there discriminatory" well if it was up to an algorithm an all-white school would stay all-white. After all there's never been a black person attending this school, so there's no reason to try to make a black kid apply. It's a feedback loop that'll keep bashing those unlikely to succeed because they're unlikely to succeed, even if they as individuals are doing everything they can. If you're constantly going to measured against what people "like you" do you'll never truly be in charge of your own destiny, which is kinda the essence of the American dream.