Artificial Intelligence on Police Activity
Approval no ANR-21-CE26-0023-01
Start: 2022
End: 2026

"Intelligent" video surveillance, facial recognition, predictive mapping, etc. are all devices that renew police action. They are supposed to build “safe cities” / “safe cities” thanks to the contributions of artificial intelligence (AI). Three main questions articulate the research:

1) Following the hypothesis of a "scientification" of police knowledge, we will wonder how algorithmic processes change the modes of knowledge and representation of crime phenomena: do we go from a police of information to "big data policing"?

2) Do these tools lead public security policies towards predictive action or are they above all tools for rationalizing police activity (by providing the police hierarchy with more precise indicators on the action field police officers)?

3) Finally, it is a question of analyzing how the debate on AI tools for urban security purposes – raising strong issues of public freedoms and the protection of personal data – pushes researchers and IT developers to integrate political and social issues into their scientific and technical productions.


This research project is based on the observation that the "algorithmization of social life", i.e. the algorithmization of an increasingly large number of social activities, significantly reconfigures the methods by which our societies produce knowledge about themselves and govern themselves. The police are not immune to these digital innovations and the changes in work they generate. This research project focuses on the ways in which the recent revival of artificial intelligence techniques (AI, and more specifically deep learning working with neural networks) is transforming both the ways of knowing the phenomena of deviance and prevention and safety measures. (Big) cities appear to be privileged testing grounds for testing and deploying these innovations within so-called “safe city / safe city” systems. An exploratory study of these projects makes it possible to distinguish two main objectives: to acquire real-time knowledge of mobility and events in a territory (movement, sound, olfactory sensors, etc., "intelligent" video surveillance, facial recognition, etc. .); develop predictive action capabilities (predictive mapping, detection of weak signals on digital social networks, calculation of the risk of occurrence of an incident, detection of crowd movements, etc.). If you have to stick to distance from the promises and marketing aims of this type of innovation, they offer the researcher circumscribed cases to examine the broader dynamics of innovation. Alongside the classic players in the field of security, such as police organizations, public authorities and private security companies, "safe city" projects bring to light less usual players: computer research laboratories, digital technologies and urban service companies. It is at the crossroads of these professional fields that we can grasp what AI does to police activity.


This project pursues three research objectives on how AI innovations affect urban safety.

  • From information policing to big data policing: towards a new quantification regime?
  • The “safe city”: predictive action or a tool for rationalizing police work?
  • The reflexive potential of the algorithmization of police activity

To carry out this research, the theoretical framework crosses the sociology of the police and surveillance studies on the one hand, and the sociology of science and technology and computer science on the other.


The empirical work will be based on a Canadian-French comparative study of four “safe city” projects: Toronto, Montreal, Paris and Nice. These four cases have three common characteristics that allow comparison: presence and mobilization of the community of AI researchers; technical appropriations and developments in this field by security, urban planning and new technology manufacturers; orders from the public authorities to renew their prevention and safety systems. It is indeed important to hold together scientific production, its technical and professional applications and the urban projects they arouse. In addition, these projects are at different stages of progress: being able to grasp them during their development will make elements that usually remain relatively invisible more accessible. In terms of survey techniques, mixed methods will be used, combining social sciences and computer science.

Role of the school and IMT's partner schools

The project coordinator will devote 85% of his time to research (field survey in France and Canada, and scientometric survey and network analysis). This project will allow him to develop his research themes, by initiating collaborations with the members of his department, and by contributing to the interdisciplinary dialogue within IMT Atlantique. Stéphanie Tillement (SGG department) also participates in the sociological part of the study, Lina Fahed (Lussi) provides the IT part of the project, with the support of Philippe Lenca (Lussi).

French and foreign academic partners

The Center for Research on Criminal Law and Institutions (Cesdip: CNRS, Versailles Saint-Quentin University, Ministry of Justice, Cergy Paris University) develops work on issues of public security, police managerial reforms, comparison of police or local management of urban disorders or even police-population relations. The center also works on city policy and urban renewal, which articulate the issues of social cohesion, urban planning and crime prevention. Jacques de Maillard and Renaud Epstein, professors of political science and researchers at the centre, will thus participate in field surveys in France and Canada and in valuation.

The Cresppa-CSU (UMR 7217), at the crossroads of sociology and political science, explores in particular questions around the city and urban segregation. Myrtille Picaud, CNRS researcher and member of the CSU, participates in the survey and field analysis, focusing on the struggles and controversies that contribute to the production of digital security policies for urban spaces.

The University of Montreal (and its International Center for Comparative Criminology) brings its expertise, its academic networks and in the security community, essential for access to Canadian cases.

Expected results

In collaborations between sociology and computer science (and more broadly between SHS and technical sciences), the first is often confined to issues of social acceptability of innovations resulting from the second. If these questions are legitimate, those maintain a utilitarian relationship between disciplines that limit the scientific impacts of interdisciplinarity. The IAAP project, while being attentive to these issues, proposes a step aside insofar as it questions the social and technical conditions of the possibility of the development of an algorithmic police, and its effects. For public authorities (police and communities): better explainability of the AI models used will allow more reflective and more reasoned use of these tools. The Franco-Canadian comparison will make it possible to report on “good practices” on both sides of the Atlantic. Recommendations for use will be made by the research team. The results will also feed the reflection of the public authorities on the regulation of these tools. In addition to the issues of open data, the police uses of AI tools inevitably raise issues in terms of the protection of personal data, transparency and fairness in the choices made by algorithms and public freedoms.

Next steps

A study trip is planned by the project coordinator for the first half of 2023 in Canada. It will be hosted at the International Center for Comparative Criminology, at the University of Montreal.

Project news

Il n'y a pas d'actualité en ce moment.