The question of bias in artificial intelligence (AI) algorithms is generally thought to be overcome by ensuring that the data set used to train the algorithm is representative of the population at large. However, Naomi Aaronson, executive director of clinical evaluation at the Blue Cross Blue Shield Association, said its not that simple because demographic data can combine in unpredictable ways and thus “the only real understanding of whether it works is in the clinical validation” of the algorithm in various settings.