Introducing Positive Discrimination in Predictive Models - Discrimination and Privacy in the Information Society - page 260

In the third and most involved method we introduced a latent variable reflecting
the actual class of a person that should have been assigned if there were no dis-
crimination. This actual non-discriminatory class is assumed to be independent of
the sensitive attribute, and the non-discriminatory labels are assumed to be discri-
minated uniformly at random, resulting in the actual labels in the data-set. The
probabilities in this model are learned using the expectation maximization algo-
rithm. We provide ways to incorporate knowledge about the discrimination
process into this algorithm. In experiments, this method unfortunately performed
poorly due to problems in the behavior of the expectation maximization algorithm.
We ended with a discussion on the positive discrimination introduced by dis-
crimination-aware data-mining and why we believe it is a better option than blind-
ly applying a discriminating off-the-shelf data-mining procedure.
References
Bishop, C.M.: Pattern Recognition and Machine Learning. Springer (2006)
Calders, T., Kamiran, F., Pechenizkiy, M.: Building classifiers with independency con-
straints. In: IEEE ICDM Workshop on Domain Driven Data Mining, pp. 13-18. IEEE
press (2009)
Calders, T., Verwer, S.: Three naive Bayes approaches for discrimination-free classifica-
tion. Data Mining and Knowledge Discovery 21(2), 277-292 (2010)
Dimitriadou, E., Hornik, K., Leisch, F., Meyer, D., Weingessel, A.: e1071: Misc functions
of the Department of Statistics. TU Wien, R package version 1 (2008)
Kamiran, F., Calders, T.: Classifying without discriminating. In: Proc. IEEE International
Conference on Computer, Control and Communication (IC4), pp. 1-6. IEEE press
(2009)
Lachiche, N., Flach, P.: Improving accuracy and cost of two-class and multi-class probabil-
istic classifiers using ROC curves. In: Proc. International Conference on Machine
Learning (ICML), pp. 416-423. AAAI Press (2003)
Langley, P., Iba, W., Thompson, K.: An analysis of Bayesian classifiers. In: Proc. Confe-
rence on Artificial Intelligence (AAAI), pp. 223-228 (1992)
Pedreschi, D., Ruggieri, S., Turini, F.: Discrimination-aware data mining. In: Proc. ACM
SIGKDD International Conference on Knowledge Discovery and Data Mining, pp.
560-568 (2008)