Soul of a new machine: Self-learning algorithms in public administration

Big data sets in conjunction with self-learning algorithms are becoming increasingly important in public administration. A growing body of literature demonstrates that the use of such technologies poses fundamental questions about the way in which predictions are generated, and the extent to which such predictions may be used in policy making. Complementing other recent works, the goal of this article is to open the machine's black box to understand and critically examine how self-learning algorithms gain agency by transforming raw data into policy recommendations that are then used by policy makers. I identify five major concerns and discuss the implications for policy making.

Automated summary

The increasing importance of big data sets and self-learning algorithms in public administration raises fundamental questions about predictions generation and their use in policy making. This article aims to open the black box of machines to understand how algorithms transform raw data into policy recommendations, identifying five major concerns and discussing their implications for policy making.