It’s no secret that Google tracks its users’ activity in order to offer better ad targeting (although the algorithms that actually process that data in order to select ads are highly guarded). But new research suggests a problem with those algorithms: that they may be sexist.
In order to gather data, a team from Carnegie Mellon University and the International Computer Science Institute built a program called AdFisher to simulate web browsing habits. They found that when these fake users were male and visited job sites, they were far more likely to be shown ads for high-paying executive jobs than ostensibly female users were.
“I think our findings suggest that there are parts of the ad ecosystem where kinds of discrimination are beginning to emerge and there is a lack of transparency,” Anupam Datta, a CMU associate professor who helped develop AdFisher, told the Massachusetts Institute of Technology’s Technology Review earlier this month. “This is concerning from a societal standpoint.”
This isn’t the first time Google has drawn ire for the way it gathers data about users, often without their knowledge. The company may be known primarily as a search engine, and it is of course a giant in that arena; Google owns between 65% and 70% of the global search engine market. But Google’s primary source of income is actually the ads it sells, with its search engine serving as a starting point for tracking users and as a platform for displaying ads.
The study does explain that the apparent sexism visible in ad display disparities may not actually be built into Google’s ad-targeting algorithms. Advertisers may be at fault in some cases, since they have options when it comes to who sees their ads.
And, of course, it’s also possible that neither Google nor advertisers are being intentional in showing more high-paying job ads to male users. As the researchers note in the study, “large-scale machine learning can behave in unexpected ways.”
For that reason, the study stops short of accusing Google of sexism, but does suggest instead that the company has “lost control” over the ad targeting system in ways that may have significant social impacts.