Carnegie Mellon built an automated testing rig called AdFisher to investigate the targeting of ads served by Google on third-party websites. It used the rig pretending to be a series of male and female job seekers. Using 17,370 simulated profiles set up for the sole purpose of visiting job sites, the researchers examined the 600,000 adverts they were shown.
Part of the research indicated that Google displayed adverts for a career coaching service for $200,000 executive jobs 1,852 times to the male group, while it was shown 318 times to the female group.
The authors said: “In particular, we found that males were shown ads encouraging the seeking of coaching services for high paying jobs more than females.”
Google’s ad targeting system takes into account different factors about personal information, browsing history and internet activity. What flagged up concern to the researchers was that the fake users had started anew with fresh profiles and behaved in exactly the same way – so gender was the only distinguishing factor.
Associate professor of computer science, Anupam Datta, who helped develop AdFisher, said it suggested that gender discrimination was ingrained in the process. However, it didn’t show where the responsibility lay – whether it was the preference of advertisers or an unintended consequence of algorithms that drive online recommendation engines.
“Many important decisions about the ads we see are being made by online systems. Oversight of these ‘black boxes’ is necessary to make sure they don’t compromise our values,” he explained.
Read more about Google:
- Google’s algorithm change could have a big impact on your website
- Google holds its hands up in bid to win over European publishers
- One year on, the right to be forgotten by Google is among the biggest policy debates of our time
The authors also said that the gender discrimination reflected here was hard to attribute to one particular factor, both due to Google’s complex profiling systems and due to the way advertisers buy and target their adverts using the search engine.
A PhD alumnus of Carnegie Mellon, Michael Tschantz said that while “we can’t look inside the black box that makes the decisions”, AdFisher can “find changes in preferences and changes in the behaviour of its virtual users that cause changes in the ads users receive”.
A Google spokeswoman said that advertisers can “choose to target the audience they want to reach”, and that the company had policies that “guide the type of interest-based ads that are allowed”.
The study suggested the selection of ads meant male workers would get more encouragement to seek coaching services for high-paying jobs which could “further the current gender pay gap”. It added that even if the result was purely down to economic reasons “it would continue to be discrimination”.
Looking into other aspects of Google ads, the researches found the site’s ad choices, which lets users actively remove certain interests from the tracking profiles, had the desired effect. Removing online dating interests for example, meant that the relevant ads stopped appearing within the top five ads shown to the test group.
Share this story