Handling Conditional Discrimination

By Indrė Žliobaitė, Faisal Kamiran, Toon Calders. 2011

194 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Causal Inference & Discovery, 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 Artificial Intelligence

In Causal Inference & Discovery

In AI Ethics, Fairness & Society

Citations in the corpus

European Union Regulations on Algorithmic Decision-Making and a "Right to Explanation" (2016, methodology): “the issue of discrimination within algorithms by introducing tools to both identify (Berendt & Preibusch, 2012; Sandvig et al., 2014) and rectify (Calders & Verwer, 2010; Hajian et al., 2011; Zliobaite et al., 2011; Berendt & Preibusch, 2014; Dive & Khedkar, 2014; Feldman et al., 201”
Causal Modeling-Based Discrimination Discovery and Removal: Criteria, Bounds, and Algorithms (2019, methodology): “Data preprocessingmethods [2], [5], [6], [17]modify the historical data to remove discriminatory patterns.”
Wasserstein Fair Classification (2019, methodology): “…that do not contain sensitive information during training (Beutel et al., 2017; Calders et al., 2009; Calmon et al., 2017; Edwards and Storkey, 2016; Feldman et al., 2015; Fish et al., 2015; Kamiran and Calders, 2009, 2012; Louizos et al., 2016; Zemel et al., 2013; ˇZliobaite et al., 2011).”
Achieving Non-Discrimination in Data Release (2016, methodology): “Their method, although in the right direction, has two signiﬁcant limitations: 1) it does not distinguish whether a partition is meaningful for measuring and removing discrimination; 2) it does not consider the situation where there exist multiple meaningful partitions.”
Learning Fair Classifiers (2015, methodology): “We observe that: i) the performance of our classifiers (C-LR and C-SVM) and the regularized logistic regression (R-LR) is comparable, ours are slightly better for the Adult data (left column) while slightly worse for the Bank data (right column), however, the regularized logistic regression uses the”
Exposing the probabilistic causal structure of discrimination (2015, background): “[45] focus on the concept of genuine requirement to detect that part of discrimination which may be explained by other, legally grounded, attributes.”
Using Loglinear Model for Discrimination Discovery and Prevention (2016, methodology): “Data preprocessing methods [2], [5] modify the historic data to remove discriminatory patterns according to some discrimination measure before learning a prediction model.”
Fairness Constraints: Mechanisms for Fair Classification (2015, methodology): “ding disparate impact in outcomes by using information about a subjects’ sensitive attributes in the decision process would constitute disparate treatment (also referred to as reverse discrimination (Zliobaite et al., 2011)). In our work, we aim to design classiﬁers that avoid both disparate impact ”
Discovering Unwarranted Associations in Data-Driven Applications with the FairTest Testing Toolkit (2015, methodology): “, favored and disfavored status [7, 16, 20, 30, 31, 38, 42, 47, 48, 56, 57], or are instead difficult to test using a sample [12]; (2) They ignore disparities in user subpopulations [7, 16, 30–32, 57]; (3) They fail to consider reasonable explanations that may negate the significance of an associati”
Discrimination Prevention using Privacy Preserving Techniques (2015, background): “There are some other issues related to DADM such as imbalanced datasets [15] and conditional discrimination [16].”
Combating discrimination using Bayesian networks (2014, methodology): “We do follow the class relabeling technique in step 2 as suggested by prior work (Calders and Verwer 2010; Zliobaite et al. 2011; Kamiran and Calders 2011; Mancuhan and Clifton 2012).”
Decoupled classifiers for fair and efficient machine learning (2017, methodology): “To our knowledge, the earliest work on trying to learn fair classifiers from historically biased data is by Pedreshi et al. (2008); see also (Zliobaite et al., 2011) and (Kamishima et al., 2011).”
Preventing undesirable behavior of intelligent machines (2019, methodology): “Similarly, for decision trees, constraining the splitting criteria using T can improve demographic parity [123], as can balancing, removing, and repeating some training data, and flipping output labels of feature vectors close to the decision tree’s decision boundaries, which introduces noise around”
Decision Theory for Discrimination-Aware Classification (2012, methodology): “This strategy is based on the discrimination model of [7] that discrimination impacts instances close to the decision boundary.”