Google’s algorithms advertise higher paying jobs to more men than women

Researchers from Carnegie Mellon university and the global Computer Science Institute investigated how Google’s ad targeting on third party websites worked, and discovered that the targeted ad system discriminates against women, showing them different job ads than men.

The study also claims Google’s transparency tool called “Ad Settings” isn’t divulging all the potentially sensitive information it uses for targeting ads.

Anupam Datta, an associate professor at Carnegie Mellon and co-author of the study, who also helped in developing the AdFisher told MIT Technology Review, ” this is a concern from a society standout point of view, because the targeted ads shown to the user some time leads them to make decisions by getting influenced by those sort of ads”. “This is a concerning from a societal standpoint”.

AdFisher works by acting as thousands of web users, each taking a carefully chosen route across the internet in such a way that an ad-targeting network like Google Ads will infer certain interests and characteristics from them.

The researchers created 17,370 fake user profiles, which they used to visit jobseeker sites, and returned with 600,000 adverts that were analysed.

“We can not determine who caused these findings due to our limited visibility into the ad ecosystem, which includes Google, advertisers, websites, and users”, said the researchers in their paper’s abstract.

Google did not officially respond to the researchers about their findings previous year. However, this June the team noticed that Google had added a disclaimer to its ad settings page.

“Advertisers can choose to target the audience they want to reach, and we have policies that guide the type of interest-based ads that are allowed”, said Google spokesperson, Andrea Faville. It can be set to act as a man or woman, then flag any differences in the adverts it is shown. “What we need now is infrastructure and tools to study these systems at much larger scale”.

AdFisher creates hundreds of simulated users, enabling researchers to run browser-based experiments in which they can identify various effects from changes in preferences or online behavior.