Research
My PhD sits between computer science, AI and education: I build AI-powered learning analytics dashboards that help K-12 teachers support what their students do outside the classroom. That makes the work inherently pluridisciplinary and what makes it so interesting.
Research Landscape
The four themes I work across, and how my papers connect them.
Learning Analytics
Turning traces of learning activity into indicators teachers can read, trust and act on.
- Dashboards Teacher-facing dashboards for K-12, from indicator selection to visual grammar.
- Indicators & Metrics Which metrics actually inform a pedagogical decision, and which only look useful.
- Co-designing Dashboards RJC EIAH 2026An approach combining prototype validation and chatbot exploration to co-design Learning Analytics Dashboards with teachers.
Human-AI Interaction
Designing with teachers and learners rather than for them, and studying what they do with AI.
- Participatory Co-design Prototypes and chatbots as elicitation devices to surface teachers' real needs.
- Teacher Decisions What teachers do with what a dashboard tells them, and what it takes to act on it.
AI in Education
Assessing what generative models can and cannot do when they grade, tutor or recommend.
- Large Language Models Prompting, architecture and evaluation choices, and how much each one really moves the needle.
- Multi-Agent Systems Orchestrating specialised agents to answer to specific use cases.
- Automatic Short Answer Grading Grading open-ended answers automatically, and measuring the gap with human graders.
- MAESTRO ECTEL 2025A multi-agent system to provide teachers with AI-powered recommendations based on learning analytics.
- Mind the Gap LAK 2026A benchmarking tool for AI systems to evaluate their performance on several configurations (model, prompts, architecture).
AI & Ethics
Naming the risks of generative AI in schools and turning them into actionable strategy.
- Responsible GenAI Integration EIAH 2025A strategic action plan for the responsible integration of generative AI in educational contexts, addressing ethical risks and challenges.
- Responsible AI Fairness, transparency and learner / teacher autonomy in the specific setting of a classroom.
- Policy & Governance Translating ethical concerns into strategic action plans.
Approach
Design-Based Research
Iterating between design, theory and deployment in ecological contexts, so each experiment feeds both the tool, scientific contributions and the model behind it.
Mixed methods
Interaction traces and benchmarks alongside focus groups: a mix of qualitative and quantitative methods.
Built, not only described
Prototypes are the research instrument: dashboards, multi-agent systems and grading benchmarks are deployed in the wild.
Context
PhD student at the IRIT lab in Toulouse, in the TALENT team, in collaboration with Kosmos Education. Field work happens in French K-12 schools, currently around Konsolidation, an out-of-class study app.