Algorithmic Discrimination against Employees

Abstract

Discrimination against candidates, employees, and platform workers is prohibited. In recent years, employers have increasingly relied on automated systems in employment relationships, and the outputs generated by these systems (including decisions, recommendations to employers, and others) may infringe on the principle of equal treatment. When automated systems produce discriminatory outcomes or base their outputs on prohibited grounds of discrimination, algorithmic discrimination arises. This article aims to formulate a definition of algorithmic discrimination in employment relationships and to analyse its forms of occurrence in the context of anti-discrimination law. The object of the study is the forms of occurrence of algorithmic discrimination in employment relationships and its assessment under international, European Union, and national legal acts establishing the prohibition of discrimination, as well as relevant case law. The article formulates and presents a definition of algorithmic discrimination, linking it to the generation of discriminatory outputs by automated systems, which result in negative consequences for protected groups distinguished by prohibited discrimination grounds. It examines forms of direct and indirect algorithmic discrimination in the context of employment relations. The study concludes that direct algorithmic discrimination arises when an automated system generates outputs based on prohibited discrimination grounds (for example, sex, age, and other aspects), whereas indirect algorithmic discrimination occurs when an apparently neutral practice (such as an algorithmic score and the criteria determining it) produces a disproportionately adverse effect on a group distinguished on prohibited discrimination grounds, and such an effect cannot be objectively justified.