In Quebec, the deployment of artificial intelligence (AI) raises growing issues of transparency, ethics, and accountability. Although Bill 25 now regulates the governance of automated systems, the implementation of explainable artificial intelligence (XIA) remains difficult to achieve in industrial settings. Current approaches focus on the technical aspects of algorithms, without offering a clear methodology for identifying and formalizing the real needs of stakeholders. This research aims to bridge this gap by proposing a methodology for identifying, structuring, and prioritizing XAI requirements, constraints, and needs, adapted to the Quebec context and aligned with the principles of trustworthy AI. It is structured around four complementary articles exploring the legal, ethical, organizational, and practical dimensions of XAI. The first three articles lay the conceptual and normative foundations necessary for the design of explainable AI: interpretation of the Quebec legal framework, operationalization of ethical principles, and characterization of stakeholders. The fourth article consolidates these findings into a comprehensive methodology, XAIRS (Explainable Artificial Intelligence Requirements Specification), a tool designed to guide the specification of XAI requirements throughout the life cycle of AI systems. This research highlights that XAI can be approached as an engineering issue, where the formalization of needs and constraints becomes a lever for transparency and accountability in the design of AI systems. The results demonstrate the feasibility and relevance of XAIRS in Quebec industrial contexts.
| Date | 15 May 2026 |
|---|
| Original language | French |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Sylvie Ratté (Supervisor) & Marc-Kevin Daoust (Co-supervisor) |
|---|
Raymond, C. (Author),
Ratté (Supervisor) &
Daoust (Co-supervisor),
15 May 2026Student thesis: Master's thesis › Master in Engineering: Engineering