Does mitigating ML's impact disparity require treatment disparity?

By Zachary Chase Lipton, Julian McAuley, Alexandra Chouldechova. 2017

225 citations recorded by the source corpus.

Subfields: AI Ethics, Fairness & Society

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In AI Ethics, Fairness & Society

Citations in the corpus

Prediction-Based Decisions and Fairness: A Catalogue of Choices, Assumptions, and Definitions (2018, background): “[80] note that adding Y (c− 1) and negating gives the cost-sensitive risk that Menon and Williamson aim to minimize [86]:”
Fairness and Abstraction in Sociotechnical Systems (2019, background): “That the mathematical definitions eliminate these nuances is a fact known to many fair-ML researchers [14, 24, 31, 53].”
Fairness Under Unawareness: Assessing Disparity When Protected Class Is Unobserved (2018, background): “, January 29–31, 2019, Atlanta, GA, USA than the disadvantaged group b, i.e. the disadvantaged group experiences an outcome disparity. Demographic disparity is simple to understand and is widely used [25, 30, 32], despite its flaws [17, 21]. When the labels for the protected class A are known, the d”
Predictive Multiplicity in Classification (2019, neutral): “This type of multiplicity leads to irreconcilable differences between subsets of predictions, which is similar to the impossibility results in the fair machine learning literature [ e.g., 11, 12, 25, 29].”
Troubling Trends in Machine Learning Scholarship (2018, background): “[49] discusses how the recent literature on fairness in ML often overloads terminology borrowed from complex legal doctrine, such as disparate impact, to name simple equations expressing particular notions of statistical parity.”
Fairness GAN (2018, background): “Reference [26] states that “the use of group-conditional decision making systems is often prohibited” whereas [27] “call[s] this premise into question.”
Explaining The Efficacy of Counterfactually-Augmented Data (2020, background): “Concerns about bias in algorithms stem in part from principles of procedural fairness (Blader & Tyler, 2003; Miller, 2017; Grgic-Hlaca et al., 2018; Lipton et al., 2018).”
On the Legal Compatibility of Fairness Definitions (2019, background): “chised and vulnerable groups, even if doing so requires unequal treatment of a particular group. Although debiasing techniques developed by the ML community might have the effect of anti-subordination[32], this balancing of the scales of historical injustice and hierarchies is rarely explicitly ackn”
Repairing without Retraining: Avoiding Disparate Impact with Counterfactual Distributions (2019, methodology): “Proposition 1 complements a recent set of impossibility results on inevitable trade-oﬀs between groups (see e.g., Lipton et al., 2018; Chouldechova, 2017; Kleinberg et al., 2016; Pleiss et al., 2017), and provides a suﬃcient testable condition that can indicate when to train diﬀerent classiﬁers for…”
Avoiding Resentment Via Monotonic Fairness (2019, background): “[31] study concepts of impact disparity and treatment disparity which overlap our own.”
Toward a better trade-off between performance and fairness with kernel-based distribution matching (2019, background): “This growing collective awareness has resulted in signiﬁcant interest to develop and popularize fairness metrics [7, 10, 13, 15, 5, 14], as well as to ﬁnd efﬁcient mitigation techniques [1, 20, 11, 8, 2, 16].”
Noise-tolerant fair classification (2019, result): “This is in line with some of the literature, since it may be forbidden to use A at test time for legal reasons (Lipton et al., 2018).”
Addressing multiple metrics of group fairness in data-driven decision making (2020, methodology): “This dataset has been used in the following studies on algorithmic fairness: [6,22] These datasets diﬀer from other datasets used in binary classiﬁcation due to the fact that they are cases of decision making processes which can be automated using machine learning.”
Fairness in Machine Learning: A Survey (2020, unknown): “While this is common in fairness research, we discuss the need to include these methodologies in fairness pipelines in Section 6.5.”
This Thing Called Fairness (2019, background): “and McAuley have suggested that fixing disparities in impact may inherently require creating disparities in treatment [97].”