Machine-learning algorithms in regulatory practice: Nine organisational challenges for regulatory agencies

Abstract

A growing body of literature discusses the impact of machine-learning algorithms on regulatory processes. This paper contributes to the predomi- nantly legal and technological literature by using a sociological-institutional perspective to identify nine organisational challenges for using algorithms in regulatory practice. Firstly, this paper identifies three forms of algorithms and regulation: regulation of algorithms, regulation through algorithms, and regulation of algorithms through algorithms. Secondly, we identify nine organisational challenges for regulation of and through algorithms based on literature analysis and empirical examples from Dutch regulatory agencies. Finally, we indicate what kind of institutional work regulatory agencies need to carry out to overcome the challenges and to develop an algorithmic regulatory practice, which calls for future empirical research.