Auditing Black-box Models by Obscuring Features

@article{Adler2016AuditingBM,
  title={Auditing Black-box Models by Obscuring Features},
  author={Philip Adler and Casey Falk and Sorelle A. Friedler and Gabriel Rybeck and Carlos Eduardo Scheidegger and Brandon Smith and Suresh Venkatasubramanian},
  journal={CoRR},
  year={2016},
  volume={abs/1602.07043}
}
Data-trained predictive models are widely used to assist in decision making. But they are used as black boxes that output a prediction or score. It is therefore hard to acquire a deeper understanding of model behavior: and in particular how different attributes influence the model prediction. This is very important when trying to interpet the behavior of complex models, or ensure that certain problematic attributes (like race or gender) are not unduly influencing decisions. In this paper, we… CONTINUE READING
Highly Cited
This paper has 17 citations. REVIEW CITATIONS
Recent Discussions
This paper has been referenced on Twitter 51 times over the past 90 days. VIEW TWEETS