Accountable Algorithms: Adopting the Public Law Toolbox Outside the Realm of Public Law

Abstract

Abstract It is well known that artificial intelligence, especially machine learning, has the potential for hugely beneficial impacts on all areas of life, but also carries with it dangers such as lack of transparency, over-rigidity of decision-making, negative feedback loops and unreasonable inferences. Such dangers also have the potential to be scaled across a whole area of decisions. There is, therefore, an increasing sense that we need greater accountability of algorithmic decision-making systems and it is argued here that public law and the grounds of judicial review are a ready-made toolkit specifically designed to render decision-makers accountable and to make precisely the kinds of trade-offs we will need to make between effectiveness and efficiency on the one hand and fairness on the other. These tools can therefore provide a blueprint as we work out how to govern ADM and render it accountable, even in a private context. This can take place by using the public law toolkit in interpreting existing legislation, informing future regulation and even through the exercise of a common law supervisory jurisdiction.