Adverse Impact and Test Validation: A Practitioner's Guide to Valid and Defensible Employment Testing

By Dan A. Biddle. 2005

165 citations recorded by the source corpus.

How researchers see this paper

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

Citations in the corpus

Adaptive Sensitive Reweighting to Mitigate Bias in Fairness-aware Classification (2018, background): “xi,yi=1,i<S ∼ N ([1, 2], [5, 2; 2, 5]) xi,yi=1,i ∈S ∼ N ([2, 0], [10, 1; 1, 4]) xi,yi=0,i<S ∼ N ([0,−1], [7, 1; 1, 7]) xi,yi=0,i ∈S ∼ N ([−5, 0], [5, 1; 1, 5])”
Does mitigating ML's disparate impact require disparate treatment? (2017, background): “In [ZVGRG17], the proposed measure of disparate impact is qa/qb, following a book by Biddle on fair employment practices [Bid06], assuming that a is the disadvantaged group.”
Learning Fair Classifiers (2015, methodology): “However, since legal definitions of fairness are often provided in terms of p% rule (Biddle, 2005), we adopted it as our measure of fairness to evaluate and compare the performance of different schemes in Section 4.”
Satisfying Real-world Goals with Dataset Constraints (2016, background): “As noted by Zafar et al. [2015], fairness is sometimes speciﬁed by a proportion, such as the 80% rule in US law that certain decisions must be in favor of group B individuals at least 80% as often as in favor of group A individuals [e.g. Biddle, 2005, Vuolo and Levy, 2013].”
Blind Justice: Fairness with Encrypted Sensitive Attributes (2018, methodology): “(6), formulated as a constrained optimization problem by Zafar et al. (2017c) mimicking the p % -rule (Biddle, 2006): for any binary protected attribute z ∈ { 0 , 1 , it aims to achieve We believe that in future work, a similar MPC approach could also be used for conditions (1), (2 where x is the…”
Fair Clustering Through Fairlets (2018, methodology): “The p % -rule is a generalization of the 80% -rule advocated by US Equal Employment Opportunity Commission (Biddle, 2006) and was used in a recent paper on mechanism for fair classiﬁcation (Zafar et al. , 2017).”
Are My Deep Learning Systems Fair? An Empirical Study of Fixed-Seed Training (2021, background): “legal case to determine a system’s fairness [9], where fairness variance could have changed the outcome.”
Training Well-Generalizing Classifiers for Fairness Metrics and Other Data-Dependent Constraints (2018, background): “The set of allowed λs is exactly the covering centers, while, following Chen et al. [2017], Agarwal et al.”
Two-Player Games for Efficient Non-Convex Constrained Optimization (2018, background): “Biddle, 2005, Vuolo and Levy, 2013], and forces the resulting classifier to make at least 80% of its positive predictions on the minority population— Goh et al. [2016] and Narasimhan [2018] discuss a number of useful constraints that are formulated similarly, both on fairness and non-fairness metric”
Potential for Discrimination in Online Targeted Advertising (2018, background): “25, mimicking the popular “80%” disparate impact rule (Biddle, 2005), to declare a group over- or underrepresented.”
Does mitigating ML's impact disparity require treatment disparity? (2017, methodology): “This metric is motivated by a text on fair employment practices [Biddle, 2006], in which it is stated that cases where the ratio qb/qa is below 0.”
The Usefulness of Unit Weights in Creating Composite Scores (2007, methodology): “In a variant of this approach, SMEs might be asked to divide up 100 points across exam components to convey similar relative weighting information (see, e.g., Biddle, 2005; Janz, Hellervik, & Gilmore, 1986).”
Fairness Constraints: Mechanisms for Fair Classification (2015, methodology): “Next, as discussed in Section 1, our deﬁnition of disparate impact leverages the “80%-rule” [Biddle, 2005].”
Non-Discriminatory Machine Learning through Convex Fairness Criteria (2018, methodology): “p-rule A popular way to measure and limit discrimination, called the p-rule (Biddle 2006), is given as follows.”