Policing, Sensemaking and the Politics of Artificial Intelligence in Canada
Bibliographic record
Abstract
One of the most crucial issues of our time for social scientists is to understand how Artificial Intelligence (AI) is transforming democratic societies. Here I study Canadian police policy making in the era of AI. As the police enacts the state monopoly of legitimate violence over a given territory, the ways in which it engages with AI to enhance this poweror notand how society responds to it, are crucial dynamics illustrative of the challenges AI pose for policymakers. I introduce the concept of Police AI Technological Innovations (PAITI): the procurement or use of a new piece of capital equipment that uses algorithms and AI to enhanceand potentially transformpolice decision-making practices. The first contribution of this dissertation is to explore how police leaders and other key policy actors make sense of PAITI. With limited information or technical background in AI, police leaders are tasked with translating complex technologies in policing terms; weight accountability and budgetary considerations; assess the needs and receptivity to change of their members; and consider how various stakeholders will respond to the AI turn in policing. Furthermore, this dissertation examines how this sensemaking impacts the very principles of democratic policing: that police services obey the rule of law (not tyrants), limit interventions in people’s lives, and are ultimately accountable to citizens. PAITI risk embedding police services within urban infrastructures, where they will be less visible or accountable to citizens, but more informed on them. PAITI policy is as such central to the continuous power struggle over the future of democratic policing. In a first theoretical chapter, this dissertation develops an assumption-based model to explore how the police simplifies PAITI according to its preferences. It is rooted in political science, Science and Technology Studies, and police sociology literature on how the police traditionally approaches innovations and organizational changes. I argue police leaders facing complex decisions regarding police AI technologies make sense of them through a simplification process centred on (1) the impact of technologies on traditional policing (enhancement or transformation), and (2) the type of surveillance capacities they enhance (direct or indirect). I introduce these simplifications under the form of two distinct, complementary continuums. On a change continuum, police leaders make sense of PAITI through a simplification process centred on the impact of technologies on traditional policing. Innovations that enhance what is valued as “real” police work by making it more efficient will be more likely to be adopted than innovations that fundamentally transform the nature of police work. On a surveillance continuum, an innovation that develops police surveillance capacities in a way that is visible to the public and habilitates the police to identify individuals directly is less likely to be favoured by police leaders.This theoretical argument is developed through the case of Canadian municipal PAITI policies, in three empirical chapters. Chapter 2 studies how environmental factors influence automatic licence plate readers (ALPR) programmatic dimensions. It fleshes out interactions between sensemaking, technical capacities, and context, by contrasting the Montreal and British Columbia cases. Chapter 3 refines our knowledge of organizations sensemaking of place-based predictive policing (PP). It gives a voice to officers who do not interact with PP. Implemented in 2017, the Vancouver Police Department exemplifies how police services’ technoscientific attitudes of PP risk perpetuating historic flaws and biases of policing under a false sense of algorithmic impartiality. Chapter 4 highlights the political dimension of body-worn cameras (BWC). The chapter notably discusses the case of Toronto, where AI was a key consideration during its 2020 BWC rollout
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.010 |
| Meta-epidemiology (narrow) | 0.000 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.002 | 0.006 |
| Science and technology studies | 0.038 | 0.023 |
| Scholarly communication | 0.022 | 0.004 |
| Open science | 0.002 | 0.007 |
| Research integrity | 0.003 | 0.005 |
| Insufficient payload (model declined to judge) | 0.012 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".