Filling the Void: How Occupational Authority Emerges from Curating Learning Algorithms

Abstract

The data-driven, statistical, and opaque nature of learning algorithms makes them distinct from previous technologies. They are speculated to be consequential for work and organizing, but how they are reconfiguring work practices is not yet fully understood. Building on a 31-month ethnographic study of predictive policing, a learning algorithm to predict crime chances, we find that its implementation was associated with the unexpected growth in authority of an occupational group (“intelligence officers”). Our process analysis finds that this change in authority was triggered by the knowledge gap that emerged between the algorithmic outputs and the police work domain. Because intelligence officers could not open the blackboxed algorithm, they instead tried to bridge the knowledge gap by means of “algorithmic curation” practices, thereby offering their own predictions and backgrounding the learning algorithm. This approach to “fill the void” reinforced the knowledge gap and kept management assuming that the predictions were made by the algorithm, which increased the authority of intelligence officers over police work allocation. Our findings offer contributions to the emerging literature on the influence of algorithmic technology on work and to the literature on occupational authority. Next to this, we provide insights into studying learning algorithms in practice.