Author: Jake Silberg, James Manyika
Publisher: McKinsey & Company
Publication Year: 2019
Summary: The following article discusses 2 opportunities that artificial intelligence (AI) presents to today’s society: the opportunity to reduce the effect of human bias, as well as the opportunity to reduce bias in these algorithms themselves. While these are great opportunities for many industries in which human bias can negatively affect businesses and societies, caution must be taken around these algorithms. Some studies have found that AI algorithms can help reduce racial disparities, whereas other studies have shown that AI algorithms include the same bias that is found in humans and deploys that bias when running an algorithm, sometimes without anybody noticing that the bias exists. The article also offers a discussion around solutions to these biases in algorithms. For example, it is established that fairness needs to be defined early on in an AI project to be able to create an unbiased algorithm. Additionally, it is also discussed that not only individual fairness but also fairness between groups, especially those groups that are already treated unfairly. Lastly, the authors provide a framework of 6 ways in which data scientists and policy leaders can create more ethical AI algorithms.