In bias we trust?

Explanation methods that help users determine whether to trust machine-learning model predictions can be less accurate for disadvantaged subgroups, a new study finds.

artificial intelligence bias computer science and technology data machine learning mit schwarzman college of computing research trust