The algorithm used by Facebook for Job advertisement is still found to be gender biased

Its only been 2 years since Facebook was filed with 5 lawsuits and had to find its way out through settlement. These lawsuits were claiming that Facebook’s employment, housing and credit ads had discrimination among them. According to the researchers from the University of Southern California, the Social Media Giant, Facebook is still responsible for delivering job advertisements unjustly based on discrimination against gender.

The three researchers namely Basileal Imana, a doctoral student of University of Southern California (USC) , John Heidemann, a research professor of Computer Science at USC and Aleksandra Korolova , assistant professor of Computer Science at USC started to inspect the algorithms used for delivering job advertisements for both Facebook and LinkedIn. They published their finding in a report that had a heading saying “Auditing for Discrimination in Algorithms Delivering Job Ads”. This report is all set to be published in the Web Conference by the end of this month.

In an email sent to an online news agency, Korolova explained her research work, she added that according to the rules made by the United States, delivery of ads should be done on the basis of qualifications. In order to test this situation, Korolova along with her research partners planned to examine while factoring out the legal lawful qualification based discrimination. She further added that while controlling the qualifications for job, Facebook Introduces distortion based on gender while targeting with balanced job ads. Lastly, for LinkedIn, Korolova said that incase of LinkedIn’s algorithm no such biased effect was found by them.

While settling the 2019s lawsuits, Facebook agreed to bring a change in their advertisements based on discrimination. Hence the main aim behind this research work carried by Korolova was to check that how the advertisement algorithms for the two giants, Facebook and LinkedIn, distort their ads based on gender and it will be compared with the calculated gender division for job. In order to execute this, they compared the overall performance of two ads based on three separate categories. These categories were software engineer, delivery driver and sales associate. Secondly, an estimate was calculated for the people that will be targeted with these advertisement of Facebook and LinkedIn.

The comparison results were shocking as it was clear that in all the three categories that were under observation showed that Facebook algorithm was biased and LinkedIn’s algorithm was not. For example in the delivery driver category, 98% males were selected for the Domino’s ad whereas for an Instacart delivery for groceries, more than 50% user who were selected were women. When these findings were shared with Facebook and were asked for an explanation, no reply was sent back by them, which lead Korolova to think of several reasons that could’ve lead Facebook to take such actions.

According to Korolova, Facebook could have more access to data due to its large number of sources and this could make them to choose a better candidate for the real world skews. She further said that it could be a possibility that the equipment used for inspecting LinkedIn’s algorithm was not up to the required level but still it was evident that LinkedIn had put in efforts to address this Algorithmic Justice. While addressing the advertising companies, Korolova expressed her thoughts saying that if they really want to increase their employee diversity they should be able to carry it out to a balanced diverse population instead of letting Facebook decided to whom their ads should be targeted to.

The researchers are trying to convince that platforms including Facebook and LinkedIn should make it more feasible to verify that whether the ads are following the anti discriminatory laws or not. And if these platforms are not willing to take steps on their own, then the lawmakers should form a legislation that will lead to a greater transparency as compared to the current one.

Photo: NurPhoto via Getty Images

Read next: Facebook is trying to pay people to give their data to make the Artificial Intelligence fairer
Previous Post Next Post