Building Classifiers with Independency Constraints

By Toon Calders, Faisal Kamiran, Mykola Pechenizkiy. 2009

598 citations recorded by the source corpus.

In this paper we study the problem of classifier learning where the input data contains unjustified dependencies between some data attributes and the class label. Such cases arise for example when the training data is collected from different sources with different labeling criteria or when the data is generated by a biased decision process. When a classifier is trained directly on such data, these undesirable dependencies will carry over to the classifier’s predictions. In order to tackle this problem, we study the classification with independency constraints problem: find an accurate model for which the predictions are independent from a given binary attribute. We propose two solutions for this problem and present an empirical validation.

Subfields: Artificial Intelligence, Optimization & Training

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 Optimization & Training

Citations in the corpus

Reducing Sentiment Bias in Language Models via Counterfactual Evaluation (2019, background): “, 2016) and enforcing independence between sensitive attributes and prediction outputs (Calders et al., 2009; Lemoine et al., 2018; Jiang et al., 2019).”
Exposing and Correcting the Gender Bias in Image Captioning Datasets and Models (2019, background): “Some works [Calders et al., 2009], [Kamiran and Calders, 2009] modify/re-weight the class labels to reduce the inﬂuence of biased data.”
Does Object Recognition Work for Everyone? (2019, background): “For instance, statistical parity [5, 6, 14, 23–25, 30, 49] states that in binary-classification settings, members of different groups should have the same chance of receiving a positive prediction.”
Towards fairer datasets: filtering and balancing the distribution of the people subtree in the ImageNet hierarchy (2019, background): “Researchers have proposed multiple fairness metrics including statistical parity [12, 13, 24, 39], disparate impact [27, 80], equalized odds [34] and individual fairness [23], and analyzed the relationship between them [43, 60].”
Gender Bias in Coreference Resolution: Evaluation and Debiasing Methods (2018, background): “Various other approaches have been proposed to produce “fair” classifiers (Calders et al., 2009; Feldman et al., 2015; Misra et al., 2016).”
Certifying and Removing Disparate Impact (2014, methodology): “[2] have also previously examined one method for changing the data in which different data items are given weights and the weights are adjusted to achieve fairness.”
Fairwashing: the risk of rationalization (2019, methodology): “With respect to fairness, among the numerous existing definitions (Narayanan, 2018), we rely on the demographic parity (Calders et al., 2009) as the fairness metric.”
Data preprocessing techniques for classification without discrimination (2011, background): “We further explore the problem of discrimination in [4] where we concentrate on the case where a labeled dataset is given, and one binary attribute B which we do not want the predictions to correlate with.”
Wasserstein Fair Classification (2019, methodology): “…the data to remove bias, or in extracting representations 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;…”
FairGAN: Fairness-aware Generative Adversarial Networks (2018, background): “remove discriminatory effect before conducting predictive analytics [4], [7], [11]–[13], the in-process approaches that enforce fairness to classifiers by introducing constraints or regularization terms to the objective functions [14], [15], and the post-process approaches that directly change the p”
Adaptive Sensitive Reweighting to Mitigate Bias in Fairness-aware Classification (2018, background): “Such approaches include massaging the dataset [3, 13, 16, 24–26] by changing class labels that are identified as mislabeled due to bias and reweighting (usually heuristically) training samples so that more importance is placed on sensitive ones [3, 11, 25].”
Delayed Impact of Fair Machine Learning (2018, background): “Demographic parity and its related formulations have been considered in numerous papers [e.g. Calders et al., 2009, Zafar et al., 2017].”
On Fairness and Calibration (2017, background): “However, it has been argued that Statistical Parity may not be applicable in many scenarios [7, 12, 18, 24], as it attempts to guarantee equal representation.”
Three naive Bayes approaches for discrimination-free classification (2010, background): “In a series of recent papers (Pedreschi et al. 2008, 2009; Kamiran and Calders 2009; Calders et al. 2009), the topic of discrimination in data mining received quite some attention.”
Weakly-Supervised Disentanglement Without Compromises (2020, background): “While they observed a strong correlation between demographic parity (Calders et al., 2009; Zliobaite, 2015) and disentanglement, the applicability of their approach is limited by the fact that disentangled representations are difficult to identify without access to explicit observations of the…”
On the Fairness of Disentangled Representations (2019, background): “To evaluate the learned representations r(x) of these observations, we assume that the set of ground-truth factors of variation include both a target factor y, which we would like to predict from the learned representation, and an underlying sensitive factor s, which we want to be fair to in the sen”
The ethics of algorithms: Mapping the debate (2016, neutral): “…may be possible to direct algorithms not to consider sensitive attributes that contribute to discrimination (Barocas and Selbst, 2015), such as gender or ethnicity (Calders et al., 2009; Kamiran and Calders, 2010; Schermer, 2011), based upon the emergence of discrimination in a particular context.”