Algorithms used to winnow and find job candidates could be discriminatory
Algorithms used to winnow lists of job candidates, find potential employees on the internet, and predict job success could be aiding discrimination based on race, sex or other protected characteristics.
Bloomberg Big Law Business outlines the potential problems. The formulas could lead employers to copy the demographics of their existing workforce, leading to the exclusion of women and minorities. Groups with less of a digital footprint could be harmed when internet-scanning technology is used to find candidates. Some algorithms may screen out those likely to be injured in the workplace or unlikely to stay in the job, rather than screen based on a person’s ability to perform the job.
Yet bias suits can be difficult, the article points out. The employer may not even be aware of the elements used in the algorithm, and programmers may be unaware of which variable has discriminatory effects. And companies making the software may claim trade secret protection for their formulas.
Washington University at St. Louis law professor Pauline Kim tells Bloomberg that employers that want to use algorithms should be clear about what they want to measure. They should make sure the algorithms measure those factors, and should check whether they are having a discriminatory impact. She also says employers should keep the data.
