If the preceding chapter sought to understand the architecture of power in the digital age, the present one turns to its most tangible manifestation within law enforcement: algorithmic policing. It is here—at the intersection of code, data, and coercive authority—that abstract concerns begin to crystallize into concrete practices.[1]
Algorithmic policing is often introduced through the language of innovation. Efficiency is foregrounded. Objectivity is implied. The suggestion—sometimes explicit, often subtle—is that technology can correct the inconsistencies and biases that have long characterized human decision-making.[2] Yet, as this chapter will demonstrate, such optimism requires careful qualification.
For while algorithms may process information with remarkable speed, they do not exist outside the social and institutional contexts that shape their design. They inherit assumptions. They reflect priorities. And, at times, they reproduce the very inequities they are purported to eliminate.[3]
1.1 Meaning and Scope of Algorithmic Policing
Defining algorithmic policing is not a straightforward task. The term encompasses a range of technologies and practices, each differing in scope, complexity, and application. At its core, however, it refers to the use of computational systems—particularly those based on machine learning and statistical modeling—to assist or influence policing decisions.[4]
These systems operate by identifying patterns within data. Historical crime records, demographic information, geospatial data, and behavioral indicators are analyzed to generate insights—sometimes descriptive, sometimes predictive.[5] The outputs may inform decisions about patrol allocation, suspect identification, or risk assessment.
Yet, the scope of algorithmic policing extends beyond mere assistance. In certain contexts, algorithmic outputs acquire a form of authority. They are treated not simply as tools, but as recommendations imbued with a degree of epistemic weight.[6] This subtle shift—from aid to influence—marks a critical point of transition.
Because once algorithmic outputs begin to shape decisions, the question is no longer whether technology is being used, but how much it is being trusted.
[1]Ferguson, The Rise of Big Data Policing (2017).
[2]O’Neil, Weapons of Math Destruction (2016).
[3]Barocas & Selbst (2016).
[4]Ferguson (2017).
[5]Lum & Isaac (2016).
[6]Pasquale, The Black Box Society (2015).