Abstract
Predictive policing involves the application of statistical models to police (and other relevant) data to forecast crime, typically through spatiotemporal (place-based) or behavioural (personbased) approaches. As the practice gained momentum in the 2010s, it attracted significant criticism – particularly regarding the quality of input data, model design, and the practical application of outputs in police operations. While a substantial body of literature acknowledges the complex impacts of predictive policing for police organisations, personnel and the public (Perry et al., 2013; Brayne, 2021; Egbert and Leese, 2021), there remains an absence of research into its implementation in Western contexts: understood here as the development, integration and use of predictive analytics within police work. This thesis addresses that gap by providing a holistic and empirically grounded account of implementation processes, including the somewhat neglected issue of oversight. Utilising a mixed methods research design, it employs two approaches to thematic analysis and integrates three data sources: eleven semi-structured interviews with English-based police personnel and academics; a systematic review of twenty-six English-language empirical organisational studies; and an in-depth analysis of the West Midlands Police Ethics Committee’s archival meeting records (n=76) (2019-2025).
Four key empirically grounded contributions are made. First, this thesis extends Wessels' (2024) input-throughput-output model by synthesising the implementation of predictive policing across international and police organisational contexts. Second, this research provides a case study of the theoretical challenges of algorithmic accountability as applied to empirical understandings of predictive policing. Third, and relatedly, the study provides insight into how oversight mechanisms can address accountability issues arising during the implementation of predictive policing. Fourth, the thesis develops the concept of fragility to argue that the predictive policing process is inherently fragile. In doing so, this thesis advances both theoretical and practical understandings of how predictive policing is adapted, contested and negotiated within contemporary police organisations.