Building Classifiers with Independency Constraints

Calders, Toon; Kamiran, Faisal; Pechenizkiy, Mykola

doi:10.1109/icdmw.2009.83

Cited by 389 publications

(399 citation statements)

References 5 publications

Supporting

Mentioning

385

Contrasting

Unclassified

Order By: Relevance

“…It 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. However, proxies for protected attributes are not easy to predict or detect (Romei and Ruggieri, 2014;, particularly when algorithms access linked datasets (Barocas and Selbst, 2015).…”

Section: Unfair Outcomes Leading To Discriminationmentioning

confidence: 99%

The ethics of algorithms: Mapping the debate

et al. 2016

View full text Add to dashboard Cite

In information societies, operations, decisions and choices previously left to humans are increasingly delegated to algorithms, which may advise, if not decide, about how data should be interpreted and what actions should be taken as a result. More and more often, algorithms mediate social processes, business transactions, governmental decisions, and how we perceive, understand, and interact among ourselves and with the environment. Gaps between the design and operation of algorithms and our understanding of their ethical implications can have severe consequences affecting individuals as well as groups and whole societies. This paper makes three contributions to clarify the ethical importance of algorithmic mediation. It provides a prescriptive map to organise the debate. It reviews the current discussion of ethical aspects of algorithms. And it assesses the available literature in order to identify areas requiring further work to develop the ethics of algorithms.

show abstract

Section: Unfair Outcomes Leading To Discriminationmentioning

confidence: 99%

The ethics of algorithms: Mapping the debate

et al. 2016

View full text Add to dashboard Cite

show abstract

“…Discrimination-aware classification originally stems from [13,14] and was further explored in [3]. The input of the discrimination-aware classification problem is a labeled dataset and one or more sensitive attributes.…”

mentioning

confidence: 99%

“…The weights on the tuples can be used directly in any method based on frequency counts. This method was first proposed in [3]. 4.…”

mentioning

confidence: 99%

Data preprocessing techniques for classification without discrimination

Kamiran¹,

Calders²

2011

Knowl Inf Syst

Self Cite

890

737

View full text Add to dashboard Cite

Recently, the following Discrimination-Aware Classification Problem was introduced: Suppose we are given training data that exhibit unlawful discrimination; e.g., toward sensitive attributes such as gender or ethnicity. The task is to learn a classifier that optimizes accuracy, but does not have this discrimination in its predictions on test data. This problem is relevant in many settings, such as when the data are generated by a biased decision process or when the sensitive attribute serves as a proxy for unobserved features. In this paper, we concentrate on the case with only one binary sensitive attribute and a two-class classification problem. We first study the theoretically optimal trade-off between accuracy and non-discrimination for pure classifiers. Then, we look at algorithmic solutions that preprocess the data to remove discrimination before a classifier is learned. We survey and extend our existing data preprocessing techniques, being suppression of the sensitive attribute, massaging the dataset by changing class labels, and reweighing or resampling the data to remove discrimination without relabeling instances. These preprocessing techniques have been implemented in a modified version of Weka and we present the results of experiments on real-life data.

show abstract

“…The key limitation of these methods is their applicability to rule based classifiers only that may not be the best classifier for a given problem. In [18], [19], [20], [9], data sampling and massaging techniques are presented for removing discrimination w.r.t. a single sensitive attribute.…”

Section: Discrimination-aware Data Miningmentioning

confidence: 99%

Classifying Socially Sensitive Data Without Discrimination: An Analysis of a Crime Suspect Dataset

Kamiran

Karim

Verwer

et al. 2012

2012 IEEE 12th International Conference on Data Mining Workshops

Self Cite

View full text Add to dashboard Cite

Abstract-Social discrimination against certain sensitive groups within society (e.g., females, blacks, minorities) is prohibited by law in many countries. To prevent discrimination arising from the use of discriminatory data, recent data mining research has focused on methods for making classifiers learned over discriminatory data discrimination-aware. Most of these methods have been tested on standard classification datasets that have been tweaked for discrimination analysis rather than over actual discriminatory data. In this paper, we study discrimination-aware classification when applied to a realworld dataset of Statistics Netherlands, which is a census body in the Netherlands. Specifically, we consider the use of classifiers for predicting whether an individual is a crime suspect, or not, to support law enforcement and security agencies' decision making. Our results show that discrimination does exist in real world datasets and blind use of classifiers learned over such datasets can exacerbate the discrimination problem. We demonstrate that discrimination-aware classification methods can mitigate the discriminatory effects and that they lead to rational and legally acceptable decisions.

show abstract

Building Classifiers with Independency Constraints

Cited by 389 publications

References 5 publications

The ethics of algorithms: Mapping the debate

The ethics of algorithms: Mapping the debate

Data preprocessing techniques for classification without discrimination

Classifying Socially Sensitive Data Without Discrimination: An Analysis of a Crime Suspect Dataset

Contact Info

Product

Resources

About