The working paper AI in service of education proposes a framework: for each level, a default progression and an assessment plan, backed by a resource bank linked to the official curriculum. The teacher modifies it freely. The framework page describes the proposal and its safeguards. This article sets out how to test it.
Why a research plan
The proposal cannot be judged on paper. A prototyping phase is needed to decide which tools to build, and in what order.
Three requirements frame the work:
- A sovereign AI solution, built on local and possibly decentralised deployments, so that student data stays under institutional control.
- A free and open-source project, so that institutions can inspect it, adapt it and avoid being locked in.
- Functions that fit any curriculum, so that the framework can serve every country willing to adopt it, not only the one where it was first tested.
The WASPL initiative has already started exploring these solutions.
Five steps
1. Prototype
Already under way, on lower-secondary mathematics: the French curriculum for cycle 4, a progression for the second year of the cycle, and simulated classes to calibrate the analysis of student answers.
2. Trials in real classrooms
With partners such as a regional education authority, a teacher-training institute or an education-sciences laboratory. The actors who may resist change (teachers, inspectors, pedagogical advisers, parents, students) are involved from the design stage, not after it. The framework page lists them, with a lever for each.
3. Impact evaluation
Indicators are defined in advance, not chosen after the results:
- teachers’ preparation time;
- curriculum coverage;
- student progress;
- actual use of the tool.
4. Generalisation
To other subjects and other countries, through a generic curriculum model based on the ISCED classification.
5. Infrastructure and governance
Sovereign hosting, maintenance of the curricula, moderation of the shared resource bank, and an economic model that keeps the project sustainable.
What success depends on
The challenge is no longer mainly technical. A prototype exists and shows that the approach is feasible.
Success now depends on three things: the buy-in of the actors, shared governance, and a rigorous evaluation of the impact in classrooms. That is why the project is designed as a common good, free and open, to which teachers, researchers, institutions and countries can contribute.