Author: Joy Looney
Publisher: Data Iku
Publication Year: 2022
Summary: The following article lays out a framework for the ethical development of machine learning models. This framework contains concrete examples of actions that data scientists and companies as a whole can take to minimize the risk of their models harming marginalized groups. For example, the article introduces the idea of using materiality, the chance that an ethical issue will occur multiplied by its cost, to rank the risks in a company’s machine learning models. This practice allows companies to focus their limited resources on the most important risk areas first.