Machine learning systems need informed (i.e., human) oversight

Rely on AI to tell us who’s worthy of a loan? Some measure of human oversight is necessary for any decision-making process reliant on machine learning, according to a research team including Carroll School Information System Professors Mike Teodorescu and Gerald Kane, who refer to such oversight as “augmentation.” Their article, “Failures of Fairness in Automation Require a Deeper Understanding of Human-ML Augmentation,” appeared in the September edition of MIS Quarterly as part of a special issue on “Managing AI.”