2  Teaching Philosophy

My pedagogical approach unfolds across a wide range of courses and through my coordination of the Centre de Dépannage et d’Apprentissage (CDA). It is structured around five core pillars.

Learning by doing (project-based learning and authentic situations). My teaching rests on a simple principle: statistics and data science are learned by practicing tasks close to real-world work, rather than by stacking decontextualized techniques. In my courses, this takes the form of projects, case studies, written and oral deliverables, and a constant effort to connect statistical results to a concrete context and audience. This orientation is explicit in my teaching philosophy and in the design of STT-1100, where students take on professional roles and produce artifacts (reports, GitHub repositories, presentations).

The classroom as a collective workshop (collaboration -> autonomy). I design the classroom as a workshop: peer interaction and small-group problem solving are part of learning, with progression toward greater autonomy (more open structure, freer choices). In STT-1100, this progression is reflected in the evolution of the Challenges and in a final team project, with assessment including peer review and GitHub repository traces. In STT-4230/6230 (R for Scientists), collaboration is supported by GitHub and software development practices (documentation, version control).

Feedback as a core learning component. I prioritize frequent feedback, often formative and sometimes automated, to support progress without unnecessarily increasing assessment load. My teaching philosophy explicitly mentions the use of learnr and Quarto notebooks, as well as feedback mechanisms through R or GitHub. In STT-1100, feedback is also embedded directly into Adventures through interventions and immediate-correction questions.

Reproducibility as a language of thought. Reproducibility is not an add-on technical feature: I treat it as an intellectual skill (organization, explicitness, justification) required for applied statistics. This is implemented through the use of Quarto (documents integrating code and text), Git/GitHub (work traceability), and related assessment criteria (repository organization, commit messages).

Generative AI: a tool to frame and make visible. I integrate generative AI within a responsible framework: it is a support tool (error clarification, documentation translation, example generation), but it does not replace reasoning or the obligation to verify. In STT-1100, a course-specific GPT assistant is described as a support resource, with instructions and reminders about verification. In an advanced R course, I designed an assessment where LLM use is mandatory and assessed, with explicit criteria for transparency, anti-hallucination safeguards, and data minimization.