Oral
Fairwashing: the risk of rationalization
Ulrich AIVODJI · Hiromi Arai · Olivier Fortineau · Sébastien Gambs · Satoshi Hara · Alain Tapp

Wed Jun 12th 04:35 -- 04:40 PM @ Seaside Ballroom

Black-box explanation is the problem of explaining how a machine learning model -- whose internal logic is hidden to the auditor and generally complex -- produces its outcomes. Current approaches for solving this problem include model explanation, outcome explanation as well as model inspection. While these techniques can be beneficial by providing interpretability, they can be used in a negative manner to perform fairwashing, which we define as promoting the perception that a machine learning model respects some ethical values while it might not be the case. In particular, we demonstrate that it is possible to systematically rationalize decisions taken by an unfair black-box model using the model explanation as well as the outcome explanation approaches with a given fairness metric. Our solution, LaundryML, is based on a regularized rule list enumeration algorithm whose objective is to search for fair rule lists approximating an unfair black-box model. We empirically evaluate our rationalization technique on black-box models trained on real-world datasets and show that one can obtain rule lists with high fidelity to the black-box model while being considerably less unfair at the same time.

Author Information

Ulrich AIVODJI (UQAM)

I am currently a postdoctoral researcher at UQÀM, in Sébastien Gambs's group. My research interests are data privacy, optimization, and machine learning. My current research focuses on two aspects of machine learning, namely privacy in collaborative machine learning, and fairness of machine learning models. I earned my Ph.D. in Computer Science at LAAS-CNRS, under the supervision of Marie-José Huguet and Marc-Olivier Killijian. During my Ph.D. I worked on privacy enhancing technologies for ride-sharing. Prior to that, I received my Engineer's degree in Software Engineering from ENSA Khouribga.

Hiromi Arai (RIKEN AIP)
Olivier Fortineau (Ensta Paristech)
Sébastien Gambs (UQAM)
Satoshi Hara (Osaka University)
Alain Tapp (Université de Montréal)

Related Events (a corresponding poster, oral, or spotlight)