How AI Bias Is Affecting Health Care—And What We Can Do About It
People are biased, and people build AI, so AI are biased, too. When AI is used in hospitals to treat patients, that bias comes to health care.
For example, a 2019 paper in Science found that a commercial risk-prediction tool was less likely to refer equally sick Black people than white people to receive extra care resources. In fact, only 17.7% of patients that the algorithm assigned to get extra care were Black, but, if the algorithm were unbiased, the percentage would be much higher—46.5%.
This episode will look at how the racial disparities baked into the health care system also make their way into the AI that the health care system uses, creating a vicious cycle. Nic Terry (an expert in the intersection of health, law, and technology) and Ravi Parikh (a practicing oncologist and bioethicist) will discuss legal and ethical concerns. Michael Abramoff (an ophthalmologist, AI pioneer, and entrepreneur) will share how he’s trying to build a fairer AI
Created with support from the Gordon and Betty Moore Foundation and the Cammann Fund at Harvard University.
