Asesoría e informes

Auditing the quality of datasets used in algorithmic decision-making systems (STOA, PE 729.541 – July 2022)

Study for the Panel for the Future of Science and Technology (STOA) within the Directorate-General for Parliamentary Research Services (EPRS) of the Secretariat of the European Parliament. Authored, among others, by our researchers Iñigo de Miguel Beriain, Pilar Nicolás Jiménez (UPV/EHU), and Guillermo Lazcoz Moratinos (CIBERER-ISCIII).

Biases are commonly considered one of the most detrimental effects of artificial intelligence (AI) use. The European Union (EU) is therefore generally committed to reducing their incidence as much as possible. However, mitigating biases is not easy, for several reasons. The types of biases in AI-based systems are many and different. Detecting them is a challenging task. Nevertheless, this is achievable if we manage to: i) increase awareness in the scientific community, technology industry, among policy-makers, and the general public; ii) implement AI with explainable components validated with appropriate benchmarks; and iii) incorporate key ethical considerations in AI implementation, ensuring that the systems maximise the wellbeing and health of the entire population. Full text available here.