Mitigating Unconscious Biases by Machine Teaching and Fairness-aware Machine Learning

Abstract

Unconscious bias is a hidden preference formed from stereotypes and other beliefs that we have unconsciously acquired. Unconscious bias regarding race or gender leads to unfair judgments when evaluating others in hiring, lending, etc. It is well known that machine learning models often output unfair evaluations for humans, and fairness-aware machine learning methods have been studied to solve this problem. In this study, we propose a method to mitigate the unconscious bias of humans in order to enable them to make fair judgments. The proposed method first lets humans make (unfair) evaluations and then applies fairness-aware machine learning to the evaluation results to obtain a fair model. In order to enable the evaluator to make the same decision as the fair model, we perform machine teaching. We conducted an experiment in which subjects were asked to predict the incomes of people. The results demonstrate that the proposed method has a corrective effect on humans who make unfair evaluations.