Discrimination aware datamining
Παυλάκη, Σοφία Κ.
Aim of the thesis is to investigate the existence or non-existence of discriminatory elements in data mining methods used for decision support systems. The different types of discrimination are analysed; direct and indirect discrimination, while there is extended reference to the existing law and regulations in relation to discrimination. Mainly discrimination is related to gender, race/nationality, religion, age, disabilities and sexual preferences. Case study for indirect and direct discrimination is analysed in the second chapter, while at the third chapter there is reference to the data mining methods which tend to produce discriminatory results. Parameter α is introduced which assists in definition of a threshold to measure discrimination in decision support. As a last point, considerations are expressed concerning the fact that there is limited research for discrimination in data mining methods as well as concerning the motivation for discrimination.