NYC Passes Bill To Restrict ‘Racist, Sexist’ Hiring Software
New York City is on track to become the first city in the nation to ban automated hiring tools unless a yearly bias audit can prove that the software won’t discriminate based on an applicant’s race or gender, and would force makers of said AI tools to open up their black box algos to scrutiny.
The bill, passed by the city council in early November and would go into effect January 2023 if signed into law, would also give candidates the option of choosing an alternative process (a human) to review their job applications, according to the Associated Press.
“I believe this technology is incredibly positive but it can produce a lot of harms if there isn’t more transparency,” said Frida Polli, co-founder and CEO of New York startup Pymetrics, which has lobbied for the legislation that favors firms such as hers which publish ‘fairness audits.’
Advocates point to a 2018 Reuters report that Amazon scrapped a similar AI recruiting tool because it favored men over women.
Pymetrics, whose core product is a suite of 12 games that are based on cognitive science experiments, paid a third party company to audit their software for bias, and to see if it passed what’s colloquially known as the ‘four-fifths’ rule – an informal hiring standard in the United States according to Technology Review.
Pymetrics and Wilson decided that the auditors would focus narrowly on one specific question: Are the company’s models fair?
They based the definition of fairness on what’s colloquially known as the four-fifths rule, which has become an informal hiring standard in the United States. The Equal Employment Opportunity Commission (EEOC) released guidelines in 1978 stating that hiring procedures should select roughly the same proportion of men and women, and of people from different racial groups. Under the four-fifths rule, Kim explains, “if men were passing 100% of the time to the next step in the hiring process, women need to pass at least 80% of the time.”
If a company’s hiring tools violate the four-fifths rule, the EEOC might take a closer look at its practices. “For an employer, it’s not a bad check,” Kim says. “If employers make sure these tools are not grossly discriminatory, in all likelihood they will not draw the attention of federal regulators.”
In theory, if Pymetrics’ suite was selecting white men for jobs, the software can correct for bias by comparing game data from those men with the results of women and people from other racial groups in order to eliminate data points which don’t correlate with race or gender, but do distinguish successful employees, according to the report, which notes that Pymetrics’s system satisfies the four-fifths rule.
Despite Pymetrics meeting the four-fifths rule, Technology Review points out that the audit didn’t actually prove that the tool is free of any bias whatsoever, nor that it picks the most qualified candidate for the job.
For example, the four-fifths rule only requires people from different genders and racial groups to pass to the next round of the hiring process at roughly the same rates. An AI hiring tool could satisfy that requirement and still be wildly inconsistent at predicting how well people from different groups actually succeed in the job once they’re hired. And if a tool predicts success more accurately for men than women, for example, that would mean it isn’t actually identifying the best qualified women, so the women who are hired “may not be as successful on the job,” says Kim.
Another issue that neither the four-fifths rule nor Pymetrics’s audit addresses is intersectionality. The rule compares men with women and one racial group with another to see if they pass at the same rates, but it doesn’t compare, say, white men with Asian men or Black women. “You could have something that satisfied the four-fifths rule [for] men versus women, Blacks versus whites, but it might disguise a bias against Black women,” Kim says. -Technology Review
We have a feeling that there’s no AI in the world that will satisfy various identity groups given how many genders, races, and species are now recognized for preferential treatment in the name of nondiscrimination.