Manipulating and measuring model interpretability

Publication
arXiv preprint arXiv:1802.07810