Source research

Discrimination, artificial intelligence, and algorithmic decision-making

Frederik Zuiderveen Borgesius · 2025 · 154 citationsRead the paper

Artificial intelligence (AI) has a huge impact on our personal lives and also on our democratic society as a whole. While AI offers vast opportunities for the benefit of people, its potential to embed and perpetuate bias and discrimination remains one of the most pressing challenges deriving from its increasing use. This new study, which was prepared by Prof. Frederik Zuiderveen Borgesius for the Anti-discrimination Department of the Council of Europe, elaborates on the risks of discrimination caused by algorithmic decision-making and other types of artificial intelligence (AI).

What the full text shows

Tested in real conditions

Facial recognition error rates are significantly higher for darker-skinned females than lighter-skinned males. — up to 34.7% error rate for darker-skinned females vs 0.8% for lighter-skinned males (Image recognition software testing)
Automated translation tools exhibit a strong tendency toward male defaults, particularly in STEM fields. (Translation from 12 gender-neutral languages (e.g., Hungarian, Chinese) into English via Google Translate)
Image search results for occupations exaggerate gender stereotypes and under-represent women. (Image search results for occupations)
Facial-tracking and camera software failed to recognize dark-colored faces or incorrectly identified open eyes as closed for Asian individuals. (Hewlett Packard facial-tracking, Nikon cameras, and passport photo automation)

Setup: Literature review of multiple third-party studies (e.g., Buolamwini and Gebru, Prates et al.) and real-world software failures.
How: The author conducted a literature review and a 'quick scan' of existing research and case studies to identify discriminatory patterns in AI. The report analyzes these findings through the lens of non-discrimination and data protection law.
Limitations: The report is a 'quick scan' rather than an in-depth mapping. · Relies on secondary literature rather than original empirical testing. · Does not cover differences in regulation across all Council of Europe member States.

Summarized by Gemma 4 from the open-access full text · 9 Oct 2026. Read it