The Architecture Matrix: Tools, Black Boxes, and the Procurement of Primary Educational AI

The international community has converged on a shared diagnosis of the risks that black-box AI systems pose to transparency, equity, and human oversight in education. UNESCO’s 2026 volume The Algorithm in the Room documents these risks and calls for agency and pedagogy to come first. What remains absent is the operational standard that translates this diagnosis into testable procurement requirements. This letter draws a distinction that existing frameworks imply but have not made explicit: the difference between AI deployed as a bounded tool and AI deployed as an architecture that tracks and steers learning trajectories over time. It further distinguishes between transparent systems and opaque black-boxes. These axes yield a 2×2 matrix and a single procurement question: does the system track the child, and can it account for its decisions? In primary education, where children lack the developmental capacity to challenge algorithmic decisions, this is not a design preference but an ethical boundary. Drawing on empirical and policy evidence, we argue that black-box AI cannot serve as educational architecture. We recommend mandating transparent, teacher-governed systems for any technology that steers a child’s learning. Black-box AI can assist, it cannot govern.