Book Title: Computational Criminology: AI Applications in Forensic Science and Criminal Justice
Editors: Dr. Xavier Louis, Dr. Surbhi Girdhar, Ms. Aswathi Chandran Nair, Mr. Ravi Kumar, and Ms. Nandini Katare
Chapter: 6
DOI: https://doi.org/10.59646/704/6
Author: Sejal Taran
Abstract
Crime is not randomly distributed in space. It clusters, concentrates, and contagiously spreads across the urban landscape in ways that are both empirically predictable and theoretically intelligible. The spatial dimension of crime the patterning of criminal events across geographic locations, and the environmental, social, and situational factors that produce that patterning has been a central object of criminological inquiry since the moral cartographers of early nineteenth-century France first visualised crime as a geographic phenomenon amenable to scientific analysis[1].
This chapter traces the trajectory of spatial crime analysis from its foundations in Geographic Information Systems (GIS) to the frontier of deep learning architectures that are reshaping the field. It does so in three broad movements. The first examines the theoretical and methodological foundations of GIS-based crime analysis, reviewing the principal analytical techniques and the criminological theories that motivate them.
The second examines the emergence and current state of deep learning approaches convolutional neural networks, spatial-temporal graph convolutional networks, transformer architectures, and their hybrid variants assessing both their analytical capabilities and their limitations. The third addresses the ethical, epistemic, and governance challenges that sophisticated spatial prediction systems generate, with particular attention to feedback loops, discriminatory data, and the rights of communities subject to AI-driven spatial surveillance[2]. The overarching argument is that spatial crime analysis stands at a productive but precarious juncture. The analytical possibilities opened by deep learning particularly the capacity to model complex, non-linear spatial-temporal dependencies across multimodal data streams are significant advances over traditional GIS methods. But these possibilities carry commensurate risks of model opaqueness, discriminatory prediction, and governance failure that the field has not yet developed adequate institutional responses to address. Technical progress must be accompanied by the epistemological self-awareness and ethical accountability that the criminal justice stakes demand.[3]
[1]Keith Harries, Mapping Crime: Principle and Practice (National Institute of Justice, 1999) 1.
[2]Paul Brantingham and Patricia Brantingham, Patterns in Crime (Macmillan, 1984) 8.
[3]Lawrence Sherman, Patrick Gartin and Michael Buerger, ‘Hot Spots of Predatory Crime: Routine Activities and the Criminology of Place’ (1989) 27(1) Criminology 27, 30.