Meta Launches New AI-Powered System To Reduce Bias In Advertisements

After a year of collaborating with the Department of Justice, Meta is introducing the first phase of its Variance Reduction System (VRS), a tool to help make advertisements more equitable across the company’s platforms. 

For now, the feature will focus on housing advertisements in the United States, with plans to expand its reach to employment and credit categories across the country in the coming year. 

According to Meta, VRS uses machine-learning technology to ensure the actual audience exposed to an advertisement aligns as closely as possible with the targets set by advertisers for that particular post. 

Once an advertisement has garnered enough eyeballs, the artificial intelligence (AI) algorithm will measure the aggregate demographic distribution of the audience, comparing the people who viewed the post to the parameters selected. 

As such, over the course of a campaign, the VRS will continue noting the demographic distribution of the advertisement’s audience, working throughout to reduce the differences between the selected audience and those who actually view the post. 

In terms of housing advertisements, the system will work to prevent posts from being limited to certain ethnic or socioeconomic groups, by matching the audience exposure against data from the US Census for the most accurate results. 

Additionally, in order to protect users’ privacy, the VRS will only rely on a widely-used method of measurement—Bayesian Improved Surname Geocoding—which is publicly available information, to measure estimated race and ethnicity. 

To put users at ease, the company has added privacy enhancements in the form of differential privacy, which helps protect individuals from being re-identified within these aggregated datasets. 

“We know we cannot wait for consensus to make progress in addressing important concerns about the potential for discrimination—especially when it comes to housing, employment, and credits ads, where the enduring effects of historically unequal treatment still have the tendency to shape economic opportunities,” says the firm. 

Meta plans to lead industry-wide discussions on the topic, and work towards building systems that will make digital advertising even fairer and equitable for all, regardless of commonalities or differences. 

[via Social Media Today and Meta, cover image via Meta]