Algorithmic Discrimination as a Threat to Digital Human Rights: a Theoretical and Legal Dimension

Abstract

The article investigates the issues regarding the protection of digital human rights in the context of the rapid implementation of artificial intelligence systems and automated decision-making within the public sector. Particular emphasis is placed on the emergence of a new phenomenon —algorithmic discrimination, which poses a serious threat to the principle of equality and creates unprecedented challenges for traditional anti-discrimination mechanisms. Algorithmic discrimination is defined as a latent, unintentional form of unfair or biased treatment towards an individual or a group of individuals possessing one or more protected characteristics, which is caused by the functioning of automated decision-making systems and lies in the large-scale and self-reinforcing restriction of their rights as a result of processing historically biased source data or applying neutral proxy variables.The article identifies and analyses the key characteristics of this phenomenon: the latent nature and insufficient transparency of the operation of algorithms, which hinder the detection of biased treatment until the violations acquire a systematic character; the absence of intent to restrict the rights of a specific individual or group of individuals, as well as the predominantly indirect nature of algorithmic discrimination; the use of proxy variables, which have a neutral character but, due to existing historical biases in society, can lead to unequal treatment; intersectionality or its cross-cutting nature, caused by the combination of several protected characteristics; a significantly accelerated and large-scale character compared to the traditional understanding of discrimination, caused by the automation of decision-making, as well as the "feedback loop", whereby the operation of algorithms further amplifies the existing unfair treatment of certain protected groups within society.The article demonstrates that the opacity ofalgorithms precludes the effective application of the classical standards for the allocation of the burden of proof under Article 14 of the ECHR, as it deprives the applicant of the tools necessary to establish a causal link between the source data, algorithmic processes, and discriminatory treatment. The study also substantiates the necessity of expanding the human rights protection doctrine and adapting the approaches of the ECtHR, incorporating the concept of "explainable artificial intelligence", in order to guarantee equitable protection against algorithmic bias.