Study finds racial bias in ads displayed on Google searches

Google searches for racially associated names appear to produce ads results for those searches that reflect racial profiling, according to a recently released study [PDF] by a Harvard University professor.

Searches containing names associated with African-Americans were more likely to produce Google AdSense ads related to arrest background checks on search result pages, according to the study by the professor, Lantanya Sweeney.

First names primarily associated with black babies, such as DeShawn, Darnell, and Germaine generated ads suggestive of an arrest in 81 to 86 percent of name searches on one website that received AdSense ads, Sweeney discovered, and on another site, arrest ads appeared 92 to 95 percent of the time.

When Professor Sweeney searched for the name "Lantanya," an ad on the page suggested an arrest, but when she paid to use InstantCheckMate.com to check, there was no arrest found.

Ads like "Trevon Jones, Arrested?" appeared on the search pages whether the search subject had an actual arrest record or not.

Names associated with whites, such as Geoffrey, Jill, and Emma, generated arrest ads 23 to 29 percent of the time on one website and zero to 60 percent of the time on another, Sweeney noted.

Inconsistent results

Results, Sweeney wrote, weren't always consistent, however. For example, "Dustin," a name associated with whites, triggered arrest ads 81 percent of the time at one website and 100 percent of the time at another.

When asked to respond, Google said: "AdWords does not conduct any racial profiling. We also have a policy which states that we will not allow ads that advocate against an organization, person or group of people. It is up to individual advertisers to decide which keywords they want to choose to trigger their ads."

The keywords linked to the arrest ads were linked by Sweeney to a single advertiser, instantcheckmate.com.

Fault?

While the results of InstantCheckmate.com's approach may appear to be racial profiling, the company may not be at fault, Sweeney said.

That's because AdSense advertisers typically have several templates associated with their ads. As those templates are served up to pages, their effectiveness is determined by the number of clicks they attract. Templates that attract the most clicks will be the ones used most often.

"Did Instant Checkmate provide ad templates suggestive of arrest disproportionately to black-identifying names?" Sweeney asked. "Or, did Instant Checkmate provide roughly the same templates evenly across racially associated names, but society clicked [on] ads suggestive of arrest more often for black-identifying names?"

The Huffington Post tried to duplicate Sweeney's results by conducting its own searches, but arrest ads were no longing appearing on the search pages for "Latonya Evans," "Latisha Smith" and "Lakisha Simmons."

Subscribe to the Today in Tech Newsletter

Comments