Last month, Twitter users uncovered a disturbing example of bias on the platform: An image-detection algorithm designed to optimize photo previews was cropping out Black faces in favor of white ones. Twitter apologized for this botched algorithm, but the bug remains.
Acts of technological racism might not always be so blatant, but they are largely unavoidable. Black defendants are more likely to be unfairly sentencedor labeled as future re-offenders, not just by judges, but also by a sentencing algorithm advertised in part as a remedy to human biases. Predictive models methodically deny ailing Black and Hispanic patients’ access to treatments that are regularly distributed to less sick white patients. Examples like these abound.
These sorts of systematic, inequality-perpetuating errors in predictive technologies are commonly known as “algorithmic bias.” They are, in short, the technological manifestations of America’s anti-Black zeitgeist. They are also the focus of my doctoral research exploring the influence of machine learning and AI on identity development. Sustained, frequent exposure to biases in automated technologies undoubtedly shape the way we see ourselves and our understanding of how the world values us. And they don’t affect people of all ages equally.