The quest to fairness in algorithmic decisions: ethical, legal and technical solutions

The rise of artificial intelligence (AI) and automation has intensified ethical and legal concerns, with fairness at the core of this discourse. This doctoral dissertation examines how fairness is conceptualised, guaranteed, and operationalised in algorithmic systems impacting human lives. As AI increasingly supports or replaces human judgment, concerns about algorithmic discrimination have become central to research and policy. Rather than rigidly defining fairness, this study explores its practical implementation in decision-making contexts with significant societal impact. Employing a multi-disciplinary legal informatics approach, the dissertation synthesises legal, philosophical, and ethical literature to provide a comprehensive overview of fairness theories and debates. It introduces a novel, multi-dimensional framework integrating legal and philosophical definitions while critically analysing EU anti-discrimination laws. By evaluating their strengths and limitations, the study refines a robust framework to address algorithmic discrimination effectively. The research also examines technical bias mitigation techniques, such as fairness metrics and synthetic data, assessing their potential in reducing algorithmic discrimination. Additionally, it offers interpretative guidance on the EU Artificial Intelligence Act (AI Act), ensuring stakeholders can navigate its obligations in practice. A key outcome of this work is the Fair, Transparent, Accountable, and Legal (Fair-y-TALe) checklist, a harm-based tool designed to prevent, evaluate, and mitigate bias throughout the AI lifecycle. Aligned with the AI Act’s provisions, this checklist operationalises fairness in algorithmic decisions, aiding in identifying and addressing discriminatory harms. Through this interdisciplinary approach, the dissertation contributes to advancing fair and accountable AI systems.