By: Ken Kingery
May 14, 2019
As the buzzwords “machine learning” continue to grow in popularity, more industries are turning to computer algorithms to answer important questions, including high-stakes fields such as healthcare, finance and criminal justice. While this trend can lead to major improvements in these realms, it can also lead to major problems when the machine learning algorithm is a so-called “black box.”
A black box is a machine learning program that does not explain how it reaches its conclusions, either because it is too complicated for a human to understand or because its inner workings are proprietary. In response to concerns that these types of models may include unjust inner workings—such as racism—another growing trend is to create additional models to “explain” these black boxes.