Facebook job ads algorithm still discriminates on gender, LinkedIn not so much Two years after Facebook settled five lawsuits claiming that its employment, housing, and credit ads illegally discriminate, researchers with the University of Southern California have found that the company still serves job ads unfairly, based on gender. In a paper titled "Auditing for Discrimination in Algorithms Delivering …
And continuing to do, despite breaking US Civil Rights laws, being caught, and promising not to do it again.
Rather than a token fine - and even fines in the millions of dollars are not a deterrent for Facebook - I would love the punishment next time to be a prohibition on selling advertising.
So facebook is targeting the adverts based on the data available while Linked in is ignoring data? Assuming I am reading it right the complaint seems to be fb discriminating by showing ads to those likely to want to see it?
Or is someone smoking something a bit strong and thinking that there is no taste discrimination based on sex? And surely fb should be able to solve this little problem by setting the rabid 100+ gender people on these researchers for their discrimination of only checking male and female?
"facebook is targeting the adverts based on the data available"
My understanding is that the data available is based on current biases and promotes the furthering of existing biases. Eg. Both Driver jobs should have equal distribution since the skill/experience required is the same, but the company where drivers were already mostly male got shown to mostly male audience
> Eg. Both Driver jobs should have equal distribution since the skill/experience required is the same, but the company where drivers were already mostly male got shown to mostly male audience
With respect, that they *should* have equal distribution is your personal opinion. Skill is certainly one factor, but not the only one. Potential interest in the job is another. As someone else pointed out, if the job involves mainly evening work, that might favour the interest of men particularly after dark.
The issue is multi-faceted and really all aspects of the job should be taken into account. Remember, people advertising for jobs are interested in reached potential applicants, not satisfying the whims of a particular activist's wishes. Those two viewpoints may coincide but that is far from guaranteed.
The fact that this is Facebook muddies the waters somewhat given their other proclivities.
@jmch
"My understanding is that the data available is based on current biases and promotes the furthering of existing biases"
That sounds like a good model for advertising. To advertise to the people likely to be interested in it based on the real world as it is makes sense (to me at least).
I know it might not make sense in the abstract academic on paper and I would agree. But people do have bias otherwise we wouldnt have so much choice in a free market.
Is that what happened though? In the past sex and race were two of the criteria that advertisers were allowed by Facebook to specify for their target audience.
Showing real estate ads to users based on race got them in to trouble previously because it was a clear violation of civil rights laws.
Have they now removed those criteria? Were they used in these cases?
That, to me, would seem to be the burning question here.
> ....in a ratio that differs from the expected gender distribution for the job.
It would be interesting to see the expected gender distribution stats for the jobs in question in addition to the difference shown by Facebook so we have something to compare to.
Regardless of how discriminatory Facebook's job ad serving is, I would have expected it to track engagement by preference, which LinkedIn possible does not.
It would also be interesting to see how well engagement correlates to interest in the job, which you would have expected to be an important metric: you don't want to serve ads to people that are not interested. The suggestion of the research is that it does not or is not sufficiently aligned, but unless we can compare the two sets of figures, then it is difficult to form an opinion.
The far, FAR more likely cause is that these "biases" are in fact the natural result of following the ROI.
I missed seeing one myself, but apparently the presentations by the ad teams to the rest of the Googlers regularly resulted in mass triggerings.
Buckle up, Buttercup.
I generally only order pizza in the evening when it is dark or near dark, and I'll bet that's the case for most people and pizza delivery is mostly a nighttime activity with few or no daytime only delivery jobs available. I can see why women would be reluctant to take such a job, whereas a job delivering groceries etc. for Instacart would have plenty of daytime hours where they would feel safer.
What if more women turned off ads than men and are hence seeing fewer ads than men? This study has no way to know that. Claiming bias on the basis of just the information that these researchers can access is a reflection of bias in the study. Besides, when there are thousands of companies that post job ads on Facebook (and LinkedIn), how will a study based on the job ads posted by just two advertisers yield statistically significant results for the entire ad platform?
Total number of ads is controlled for - they're comparing the relative frequencies of two different adverts, not asserting that they should both be 50:50 - and they aren't just looking at two advertisers, they've taken several pairs of similar jobs for different industries. Three specific pairs are described in the article, and the researchers specifically highlight that they want more data, but that gathering it is costly.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
