De (On)mogelijkheden van machine learning voor het verminderen van bias en discriminatie bij personeelsbeslissingen
Annemarie M.F. Hiemstra ( Erasmus Universiteit Rotterdam)
Tatjana Cassel ( Erasmus Universiteit Rotterdam)
Marise Ph Born ( Erasmus Universiteit Rotterdam)
Cynthia C.S. Liem (TU Delft - Multimedia Computing)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
In this article, we describe the implementation of algorithms based on machine learning for personnel selection procedures and how this data-driven approach corresponds to and differentiates from classical psychological assessment. We discuss if, and in what way, bias and discrimination occur when using algorithms based on machine learning for personnel selection. For this reason, we conducted a literature review (covering 2016-2019) from which 41 articles were included. The results indicate that algorithms possibly lead to reduced (indirect) discrimination compared to some other selection methods. This is one of the reasons why the development of algorithms for personnel selection has increased quickly and the number of vendors has grown fast. It is insufficiently possible yet, however, to ascertain if the promise is kept. First, this is because algorithms are often trade secrets (lack of transparency). Second, the validity and reliability of data used for the development of algorithms are not always clear. Furthermore, psychological selection issues about diversity and validity cannot (yet) be solved by algorithms. The increasing attention for the topic, expressed by a large growth in publications, is hopeful. We conclude with recommendations for the detection and reduction of bias and discrimination when using machine learning algorithms for personnel selection.