Fair tree classifier using strong demographic parity

Machine Learning, vol. 113, pp. 3305–3324

Abstract

Abstract When dealing with sensitive data in automated data-driven decision-making, an important concern is to learn predictors with high performance towards a class label, whilst minimising for the discrimination towards any sensitive attribute, like gender or race, induced from biased data. Hybrid tree optimisation criteria have been proposed which combine classification performance and fairness. Although the threshold-free ROC-AUC is the standard for measuring classification model performance, current fair tree classification methods mainly optimise for a fixed threshold on the fairness metric. In this paper, we propose SCAFF—splitting criterion AUC for Fairness—a compound decision tree splitting criterion which combines the threshold-free strong demographic parity with ROC-AUC termed, easily applicable as an ensemble. Our method simultaneously leverages multiple sensitive attributes of which the values may be multicategorical, and is tunable with respect to the unavoidable performance-fairness trade-off. In our experiments, we demonstrate how SCAFF generates effective models with competitive performance and fairness with respect to binary, multicategorical, and multiple sensitive attributes.

Authors 4

1. António Pereira Barata corresponding Aachen

   Leiden University · Ministry of Infrastructure and Water Management

   Affiliation as printed

   ILT , Ministry of Infrastructure and Water Management , Graadt van Roggenweg 500 , 3531 AH Utrecht , The Netherlands

   LIACS , Leiden University , Niels Bohrweg 1 , 2333 CA Leiden , The Netherlands

   LIACS, Leiden University, Niels Bohrweg 1, 2333 CA, Leiden, The Netherlands

2. Frank W. Takes Aachen

   Leiden University

   Affiliation as printed

   LIACS, Leiden University, Leiden, The Netherlands

   LIACS, Leiden University, Niels Bohrweg 1, 2333 CA, Leiden, The Netherlands

3. H.J. van den Herik Aachen

   Leiden University

   Affiliation as printed

   LCDS, Leiden University, Leiden, The Netherlands

   LCDS, Leiden University, Niels Bohrweg 1, 2333 CA, Leiden, The Netherlands

4. Cor J. Veenman Aachen

   Leiden University · Netherlands Organisation for Applied Scientific Research

   Affiliation as printed

   LIACS, Leiden University, Leiden, The Netherlands

   Netherlands Organization for Applied Scientific Research (TNO), The Hague, The Netherlands

   LIACS, Leiden University, Niels Bohrweg 1, 2333 CA, Leiden, The Netherlands