LONDON: Meta has announced the rolling out of a Variance Reduction System in the US as part of its attempt to fulfill its vow to reduce discrimination through artificial intelligence.
The VRS works to ensure an ad’s real crowd better matches the target audience, meaning it should not skew toward specific demographics, reported Engadget.
When enough people have viewed an ad, a machine learning system compares the combined demographics of viewers with those the marketers intended to reach.
It then adjusts the ad’s auction value, which is its prospect to attract viewers, to set the frequency of displaying the ad to certain cultural groups. The VRS continues to operate throughout an ad run.
To protect against piracy, the system is equipped with a differential technology that creates “noise” to prevent the AI from learning individual demographic data — such as age, gender and ethnicity — over time.
Meta plans to first apply this anti-discrimination method to the housing ads. The VRS will reach credit and employment ads over the following year, the company said.
This move comes after over a year of work in cooperation with the Justice Department and the Department of Housing and Urban Development.
Meta, dubbed Facebook at the time, was charged in 2019 with enabling discrimination in housing ads by letting advertisers exclude specific demographics, including those protected by the Fair Housing Act.
In a June 2022 settlement, the social media giant said it would both deploy the VRS and scrap the “Special Ad Audience” tool whose algorithm allegedly led to discrimination.
