Fairness without Harm: Decoupled Classifiers with Preference Guarantees

In domains such as medicine, it can be acceptable for machine learning models to include sensitive attributes such as gender and ethnicity. In this work, we argue that when there is this kind of treatment disparity then it should be in the best interest of each group. Drawing on ethical principles such as beneficence (“do the best”) and non-maleficence (“do no harm”), we show how to use sensitive attributes to train decoupled classifiers that satisfy preference guarantees. These guarantees ensure the majority of individuals in each group prefer their assigned classifier to (i) a pooled model that ignores group membership (rationality), and (ii) the model assigned to any other group (envy-freeness). We introduce a recursive procedure that adaptively selects group attributes for decoupling, and present formal conditions to ensure preference guarantees in terms of generalization error. We validate the effectiveness of the procedure on real-world datasets, showing that it improves accuracy without violating preference guarantees on test data.

[1]  Nathan Srebro,et al.  Learning Non-Discriminatory Predictors , 2017, COLT.

[2]  Dan Suciu,et al.  Capuchin: Causal Database Repair for Algorithmic Fairness , 2019, ArXiv.

[3]  C. Rudin,et al.  Clinical Prediction Models for Sleep Apnea: The Importance of Medical History over Symptoms. , 2016, Journal of clinical sleep medicine : JCSM : official publication of the American Academy of Sleep Medicine.

[4]  Indre Zliobaite,et al.  A survey on measuring indirect discrimination in machine learning , 2015, ArXiv.

[5]  Cynthia Rudin,et al.  Supersparse linear integer models for optimized medical scoring systems , 2015, Machine Learning.

[6]  Krishna P. Gummadi,et al.  Fairness Constraints: Mechanisms for Fair Classification , 2015, AISTATS.

[7]  Berk Ustun,et al.  Repairing without Retraining: Avoiding Disparate Impact with Counterfactual Distributions , 2019, ICML.

[8]  Harikrishna Narasimhan,et al.  Learning with Complex Loss Functions and Constraints , 2018, AISTATS.

[9]  D. Lynch,et al.  The National Lung Screening Trial: overview and study design. , 2011, Radiology.

[10]  Erez Shmueli,et al.  Algorithmic Fairness , 2020, ArXiv.

[11]  Deborah Hellman,et al.  When Is Discrimination Wrong , 2008 .

[12]  M. Ghassemi,et al.  Can AI Help Reduce Disparities in General Medical and Mental Health Care? , 2019, AMA journal of ethics.

[13]  Sherri Rose,et al.  Fair regression for health care spending , 2019, Biometrics.

[14]  Andrew D. Selbst,et al.  Big Data's Disparate Impact , 2016 .

[15]  Kadija Ferryman,et al.  Fairness in precision medicine , 2018 .

[16]  Seth Neel,et al.  Preventing Fairness Gerrymandering: Auditing and Learning for Subgroup Fairness , 2017, ICML.

[17]  Richard J. Arneson.  What Is Wrongful Discrimination , 2006 .

[18]  Krishna P. Gummadi,et al.  Fairness Beyond Disparate Treatment & Disparate Impact: Learning Classification without Disparate Mistreatment , 2016, WWW.

[19]  Maya R. Gupta,et al.  Training Well-Generalizing Classifiers for Fairness Metrics and Other Data-Dependent Constraints , 2018, ICML.

[20]  Adam Tauman Kalai,et al.  Decoupled Classifiers for Group-Fair and Efficient Machine Learning , 2017, FAT.

[21]  Percy Liang,et al.  Fairness Without Demographics in Repeated Loss Minimization , 2018, ICML.

[22]  Sharad Goel,et al.  The Measure and Mismeasure of Fairness: A Critical Review of Fair Machine Learning , 2018, ArXiv.

[23]  O. J. Dunn.  Multiple Comparisons among Means , 1961 .

[24]  Ilya Shpitser,et al.  Fair Inference on Outcomes , 2017, AAAI.

[25]  J. Paul Brooks,et al.  Support Vector Machines with the Ramp Loss and the Hard Margin Loss , 2011, Oper. Res..

[26]  David Sontag,et al.  Why Is My Classifier Discriminatory? , 2018, NeurIPS.

[27]  S. Athar.  Principles of Biomedical Ethics , 2011, The Journal of IMA.

[28]  Matt J. Kusner,et al.  Counterfactual Fairness , 2017, NIPS.

[29]  S. Fullerton,et al.  Genomics is failing on diversity , 2016, Nature.

[30]  Cynthia Rudin,et al.  Interpretable classification models for recidivism prediction , 2015, 1503.07810.

[31]  Yiling Chen,et al.  Fair classification and social welfare , 2019, FAT*.

[32]  Scott Sanner,et al.  Algorithms for Direct 0-1 Loss Optimization in Binary Classification , 2013, ICML.

[33]  Krishna P. Gummadi,et al.  From Parity to Preference-based Notions of Fairness in Classification , 2017, NIPS.

[34]  Elias Bareinboim,et al.  Equality of Opportunity in Classification: A Causal Approach , 2018, NeurIPS.

[35]  Avi Feller,et al.  Algorithmic Decision Making and the Cost of Fairness , 2017, KDD.

[36]  K. Crawford.  The Hidden Biases in Big Data , 2013 .

[37]  E. Vayena,et al.  Machine learning in medicine: Addressing ethical challenges , 2018, PLoS medicine.

[38]  Nathan Srebro,et al.  Equality of Opportunity in Supervised Learning , 2016, NIPS.

[39]  Carlos Eduardo Scheidegger,et al.  Certifying and Removing Disparate Impact , 2014, KDD.

[40]  Adam Tauman Kalai,et al.  Unleashing Linear Optimizers for Group-Fair Learning and Optimization , 2018, COLT.

[41]  Alexandra Chouldechova,et al.  Does mitigating ML's impact disparity require treatment disparity? , 2017, NeurIPS.

[42]  John Langford,et al.  A Reductions Approach to Fair Classification , 2018, ICML.

[43]  Thomas G. Dietterich.  Approximate Statistical Tests for Comparing Supervised Classification Learning Algorithms , 1998, Neural Computation.