Azimi's Behavioral Architecture Redefines GenAI in Education

A new arXiv paper by Sepinoud Azimi argues that current AI-in-education responses—banning AI or building content-only tutors—fail because they ignore decades of research showing personality, motivation, and emotional state shape learning outcomes as strongly as cognitive ability. The paper proposes 'behavioral architecture' as an alternative, where the system adapts to how a student learns, not only to what they learn next. This shifts the designer role from the AI vendor alone to a distribution among educator, student, and system, with students co-authoring and controlling their records.

The architecture requires governance at the EU level, as the institution operating the system is the same one assessing the student, and individual institutions cannot provide the structural protections needed. Azimi names a vacancy: the designer role is currently unoccupied, and occupying it requires infrastructure that does not yet exist. Five empirical questions are proposed to test whether the architecture delivers on its claims, including whether the distributed design improves learning outcomes and student agency.