Thread by @techreview

If you’re on the job hunt, you might be hired based on your performance in computer games scored by AI. Before you start the process, there are a few things you should know. We’ll walk you through them in this thread.

More companies are using AI-based hiring tools to manage the flood of applications they receive–especially now that there are more jobless workers in the US due to the pandemic.
mercer.com/content/dam/me…

As with other AI applications, though, researchers have found that some hiring tools produce biased results. Many are now advocating for greater transparency and more regulation.

One company striving for bias-free results is Pymetrics, which many large US firms hire to screen job applicants.
pymetrics.ai

While an applicant plays their way through a series of games, an AI system measures cognitive, social, and emotional attributes such as risk tolerance and learning ability. Those who display traits desired for the role advance to the next hiring stage.

So what does this mean for you? When applying for jobs, you may be asked to use a tool like Pymetrics’s system to measure your suitability for a role.

Companies need to ensure these AI tools are not contributing to discrimination based on race or gender.

To test this, Pymetrics paid a team of computer scientists from Northeastern University to audit its AI. Theirs was one of the first third-party audits of hiring algorithms initiated by a company for its own tool.

“What Pymetrics is doing, which is bringing in a neutral, third party to audit, is a really good direction in which to be moving,” says Pauline Kim, a law professor at Washington University in St. Louis.

The audit focused on one specific question: Are the models Pymetrics produces fair, based on what’s colloquially known as the four-fifths rule?

The four-fifths rule has become an informal hiring standard in the United States. The 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.
govinfo.gov/content/pkg/CF…

Pymetrics’s system satisfies the four-fifths rule. But that doesn’t mean the tool is free of any bias or that the company’s algorithm actually picks the most qualified candidates for any job.

Another issue that the audit didn’t cover is intersectionality: “You could have something that satisfied the four-fifths rule, men versus women, blacks versus whites, but it might disguise a bias against black women,” Kim says.

Measures that could better define audits and address bias in AI more broadly are starting to pop up in the US.
This story was reported by @HilkeSchellmann, and turned into this thread by @madisonumina.

👀READ the full story on auditing AI hiring tools.
technologyreview.com/2021/02/11/101…

Want to support our mission-driven journalism? For $50 per year, you’ll get unlimited access to our journalism and our award-winning newsletter on artificial intelligence, The Algorithm.
forms.technologyreview.com/subscriptions/…