Interpretability Can Be Actionable

Hadas Orgad, Fazl Barez, Tal Haklay +9

Interpretability aims to explain the behavior of deep neural networks. Despite rapid growth, there is mounting concern that much of this work has not translated into practical impa…