Fairness and algorithmic decision-making

Abstract

The paper focuses on one of the most urgent risks of artificial intelligence, and more specifically of algorithmic decision-making (ADM), that is, the risk of being unfair. In the first section we provide an overview of the discussion on fairness in ADM and show its shortcomings; in the second section we pursue an ethical inquiry into the concept of fairness, and identify its main dimensions and components, drawing insight from a renewed reflection on respect, which goes beyond the idea of equal respect to include respect for particular individuals too. In the third section we show how our conceptual re-elaboration of fairness can help identify the criteria that ought to steer the ethical design of ADM-based systems to make them really fair.