Can Hiring Algorithms be Impartial Decision-Makers?

AI decision-makers were believed to make more equitable hiring decisions than human decision-makers, who are influenced by their own internal biases. However, AI decision-makers are not inherently fairer. They are designed, trained and implemented by imperfect humans and can perpetuate the same discrimination they were supposed to combat. There is a variety of reasons for this. One reason is that the training data is biased, one group is under- or overrepresented, often in a manner that reflects the existing inequalities in society. Another reason is that algorithms, on their own, can find correlations between certain traits, such as sex, and a prospective employee’s chances of success at the sought-after position. Or the AI decisionmakers find a proxy, a trait that is not a protected characteristic but has the same adverse impact against a group with that characteristic. This paper will focus on AI’s effect on gender discrimination in hiring decisions and how to avert or rectify AI gender discrimination.