Regulating algorithmic decision-making (ADM) challenges policy makers as applications of ADM touch upon basic rights and values. A salient issue in recent risk-based regulatory proposals is fairness, mostly vaguely linked to normative claims of justice. In this paper, we focus on two flaws in the current algorithmic fairness discourse that impede to tackle larger justice claims. The first...
Application of AI to the justice system demands that we carefully consider whether the algorithms are just. While this statement seems simple, it highlights a deep and fundamental oversight in discussions of the ethical use of AI in judicial matters. It is an understandable oversight. The focus regarding the ethics of AI use in this context often instead deal with particular algorithms or...
Fair Machine Learning (ML) research aims to provide and improve criteria for the fairness of ML algorithms. We review the proposed metrics which usually evaluate either the fairness of the distribution of goods, opportunities etc. as produced by the algorithm’s decision (outcome-based criteria) or the fairness of the process itself which is used to arrive at a decision (*procedural...