Author: Michael A. Simon
Publisher: Drug: Discovery & Development
Publication Year: 2022
Summary: The following article first explains how implicit bias is present in the medical setting, and then links this train of thought into how implicit bias can creep into modeling through the nature of artificial intelligence (AI) also. It explains how racial bias within the models created by a certain AI led to medical outcomes that were lacking for people who were Black. The article then outlines 3 actionable steps to take. The first is to gather rich longitudinal data that encompasses the full population. The second is to select models that predict outcomes that are universally acceptable and applicable or unavoidable. The third is to think critically about algorithmic outputs to be sure they are not playing into biases.