Personal autonomy as an ethical foundation for opaque algorithmic decision systems

Abstract

AI is becoming a highly efficient instrument for decision-making in relation to the distribution of goods, services or prerogatives in different public and private administrative systems. The problem is that the greatest efficiency in this area is obtained thanks to black box algorithmic systems for which, due to their technical characteristics, explanations cannot be provided of how they have made their decisions. This has led a number of scholars to actively question the use of such systems, arguing that the lack of explanations in important decisions for the subjects poses a serious threat to their autonomy and, with it, an attack on their dignity. In this article the basic idea is accepted that the opacity of these systems implies, in principle, an erosion of personal autonomy. However, it is also argued that this idea does not rule out the possibility that the lack of explanations may at times be justified. To support this thesis, we first analyze the interpretations of three basic criteria (agential, justificatory and normative) that have given rise to the aforementioned position, based on dignity, which is the object of critique here. Alternative interpretations of such criteria are then given, from which to deduce a certain flexibility in the demand for transparency in algorithmic decision-making systems. Finally, three principles are derived from this proposal to ethically regulate the use of this type of system.