Interfacing Predictive Policing: Objectivity and Expertise Between Police and Data Scientists

• In Event: STS and Computational Knowledge Production in Policing and Criminal Justice - II

Abstract

Predictive policing has piqued the interest of scholars, policy makers, and community advocates who question the validity and efficacy of this new policing technique. But how are we to understand the technology at the heart of this new strategy when the computations that generate predictions are blackboxed in proprietary algorithms? Legal scholars and data scientists have demonstrated that supposedly objective algorithms may produce biased outcomes (Lum et al 2016, Ensign et al 2017) and violate 4th amendment rights (Pasquale 2015, Ferguson 2017). These approaches put too much emphasis on the algorithm itself, and fail to recognize the full panoply of actors involved in the socio-technical assemblage of predictive policing algorithms.
This paper looks at predictive policing interfaces as epistemic boundary objects that mediate claims to expertise from data scientists and police. Building on STS approaches that follow work done with, around, and to representations (Vertesi 2015, Lynch and Woolgar 1990) this study combines observations of policing industry events with a political semiotic analysis (Wagner-Pacifici 2017) of predictive policing interfaces. In doing so, this paper will show how these interfaces are imbued with objectivity and authority through both their design and the way they are marketed in policing conferences and trade shows demonstrations. This study contributes to both STS studies of algorithms and workplace studies that follow epistemic representations within and across professional fields. Finally, this paper demonstrates the importance of considering the full social life of an algorithm to understand their impact at work.

Author

• Zoe Carey, New School for Social Research