Algorithmic Decision-Making and the Problem of Control

Abstract

In the legal sector, as well as in public policy or medicine, decisions are increasingly being delegated to learning algorithms. In this paper we argue that these delegation practices involve trade-offs in terms of control. These trade-offs will be scrutinized from a normative point of view. In particular, we will focus on two (potential) sources for loss of control: (i) epistemic dependence and (ii) the accountability gap. By drawing on the literature on testimony and moral responsibility, in addition to discussing some of the basic concepts of machine learning, we argue that relevant loss of control might shape the motivational structure of decision-makers in a way that is ethically problematic. Therefore, even under the assumption that learning algorithms make fairer or more objective decisions than human experts, associated costs stemming from the loss of control might yet make delegating high-stakes decisions to learning algorithms ethically questionable.

Ähnliche Arbeiten