Publications Search
Aarshee Mishra, Przemyslaw A. Grabowicz, Nicholas Perello
Towards Fair and Explainable Supervised Learning Proceedings Article
In: ICML Workshop on Socially Responsible Machine Learning, 2021.
Abstract | Links | BibTeX | Tags: Fairness
@inproceedings{Mishra2021,
title = {Towards Fair and Explainable Supervised Learning},
author = {Aarshee Mishra and Przemyslaw A. Grabowicz and Nicholas Perello},
url = {https://drive.google.com/file/d/1z24hITF0Xrlc_IX_rOZVZ2aigOj1hxhD/view?usp=sharing},
year = {2021},
date = {2021-01-01},
booktitle = {ICML Workshop on Socially Responsible Machine Learning},
abstract = {Algorithms that aid human decision-making may inadvertently discriminate against certain protected groups. We formalize direct discrimination as a direct causal effect of the protected attributes on the decisions, while textitinduced indirect discrimination as a change in the influence of non-protected features associated with the protected attributes. The measurements of average treatment effect (ATE) and SHapley Additive exPlanations (SHAP) reveal that state-of-the-art fair learning methods can inadvertently induce indirect discrimination in synthetic and real-world datasets. To inhibit discrimination in algorithmic systems, we propose to nullify the influence of the protected attribute on the output of the system, while preserving the influence of remaining features. To achieve this objective, we introduce a risk minimization method which optimizes for the proposed fairness objective. We show that the method leverages model accuracy and disparity measures.},
keywords = {Fairness},
pubstate = {published},
tppubtype = {inproceedings}
}
Algorithms that aid human decision-making may inadvertently discriminate against certain protected groups. We formalize direct discrimination as a direct causal effect of the protected attributes on the decisions, while textitinduced indirect discrimination as a change in the influence of non-protected features associated with the protected attributes. The measurements of average treatment effect (ATE) and SHapley Additive exPlanations (SHAP) reveal that state-of-the-art fair learning methods can inadvertently induce indirect discrimination in synthetic and real-world datasets. To inhibit discrimination in algorithmic systems, we propose to nullify the influence of the protected attribute on the output of the system, while preserving the influence of remaining features. To achieve this objective, we introduce a risk minimization method which optimizes for the proposed fairness objective. We show that the method leverages model accuracy and disparity measures.
Przemyslaw A. Grabowicz, Nicholas Perello, Kenta Takatsu
Resilience of Supervised Learning Algorithms to Discriminatory Data Perturbations Journal Article
In: 2019.
Abstract | Links | BibTeX | Tags: Fairness
@article{Grabowicz2019c,
title = {Resilience of Supervised Learning Algorithms to Discriminatory Data Perturbations},
author = {Przemyslaw A. Grabowicz and Nicholas Perello and Kenta Takatsu},
url = {http://arxiv.org/abs/1912.08189},
year = {2019},
date = {2019-12-01},
abstract = {Discrimination is a focal concern in supervised learning algorithms augmenting human decision-making. These systems are trained using historical data, which may have been tainted by discrimination, and may learn biases against the protected groups. An important question is how to train models without propagating discrimination. In this study, we i) define and model discrimination as perturbations of a data-generating process and show how discrimination can be induced via attributes correlated with the protected attributes; ii) introduce a measure of resilience of a supervised learning algorithm to potentially discriminatory data perturbations, iii) propose a novel supervised learning algorithm that inhibits discrimination, and iv) show that it is more resilient to discriminatory perturbations in synthetic and real-world datasets than state-of-the-art learning algorithms. The proposed method can be used with general supervised learning algorithms and avoids inducement of discrimination, while maximizing model accuracy.},
keywords = {Fairness},
pubstate = {published},
tppubtype = {article}
}
Discrimination is a focal concern in supervised learning algorithms augmenting human decision-making. These systems are trained using historical data, which may have been tainted by discrimination, and may learn biases against the protected groups. An important question is how to train models without propagating discrimination. In this study, we i) define and model discrimination as perturbations of a data-generating process and show how discrimination can be induced via attributes correlated with the protected attributes; ii) introduce a measure of resilience of a supervised learning algorithm to potentially discriminatory data perturbations, iii) propose a novel supervised learning algorithm that inhibits discrimination, and iv) show that it is more resilient to discriminatory perturbations in synthetic and real-world datasets than state-of-the-art learning algorithms. The proposed method can be used with general supervised learning algorithms and avoids inducement of discrimination, while maximizing model accuracy.