A computer algorithm widely used by health care providers and insurers to predict health risk was biased against black patients, according to a new study. The software was meant to determine who had the most complex medical needs based on a patient's health care bills, and therefore would be most likely to benefit from additional care. But it failed to account for the fact that black patients, regardless of health status, actually generate lower costs because they tend to have a harder time accessing care in the first place, due to structural inequities. We'll talk to the study's lead author, a UC Berkeley researcher, on what these findings mean for inequality in health care as well as bias in artificial intelligence programs.