In Loren Goodman’s most recent article in VentureBeat, he outlines the ways in which societal biases are exacerbated by artificial intelligence and machine learning models. According to Goodman, the primary problems with machine learning today are the “black box” problem and the models’ limited ability to think only within the data used to train it. The largest point of concern is that businesses often do not have a method to identify the biased practices of their models.
There are real-world costs associated with AI bias. From the growing scrutiny on social service programs to the criticisms on how some screening systems in healthcare do not take racial disparities into consideration, bias is present and can impact people’s daily lives. In Goodman’s view, the best way to combat this algorithmic bias is to integrate people into the decision making process. Offering more transparency and accountability will help consumers understand AI and the decisions it makes better.