Advocates for algorithmic fairness dispel accuracy tradeoff with new debiasing technique

Developers of artificial intelligence models used to make decisions such as mortgage eligibility can reduce the bias in such systems against minority populations without sacrificing their accuracy, or increasing their risk, according to a new report from the National Fair Housing Alliance.

The April 24 report developed in collaboration with the Los Angeles-based company FairPlay AI details a novel technique called distribution matching, which “can increase mortgage underwriting and pricing fairness in excess of 13 percent with no functional...