Algorithmic Bias in the Light of the GDPR and the Proposed AI Act
Maciej Olejnik, Wiktoria Morawska (ed.), "(In)equality. Faces of modern Europe", Wydawnictwo Centrum Studiów Niemieckich i Europejskich im. Willy’ego Brandta, Wrocław, 2022 (forthcoming)
16 Pages Posted: 27 May 2022
Date Written: May 8, 2022
Abstract
The paper discusses that the digitalization of socioeconomic life comes along with the many benefits but also generates some risks, such as algorithmic bias. European legislation is trying to address these threats. For algorithmic bias, the main solutions are the General Data Protection Regulation and the proposed Artificial Intelligence Act. The claim I defend is that these regulations are ineffective and poorly considered. I point out that the tool of human oversight, which the GDPR and the AI Act require in terms of adapting some algorithms, leads to a false sense of comfort without providing the actual legal protection. The argument proceeds in three main steps: i) I characterize the problem of algorithmic bias and its causes, focusing on the data flaws (resulting from both unrepresentative data and representative data reflecting the underlying social inequalities); ii) I focus on its consequences; iii) I discuss the relevant legal regulations and point out their main drawbacks holes. The main purpose of the article is to conduct a critical analysis of the European legal framework addressing algorithmic bias, draw attention to its shortcomings, and suggest ways forward.
Keywords: algorithmic bias, human oversight, General Data Protection Regulation (GDPR), Artificial Intelligence Act, digital discrimination
Suggested Citation: Suggested Citation