Hanbit Elementary School, Seongnam-si, Gyeonggi-do, Republic of Korea

Abstract: Since around 2025, South Korea has been advancing AI digital textbooks and the digitalization of basic education, yet a new malady has emerged in primary classrooms: the more advanced the technology, the more a child's "mistakes" are compressed into red crosses; the more teachers rely on platform dashboards, the harder it becomes to see how a child thought of 12×14 as 12×10 plus 12×4, or how, in a fraction diagram, they drew "one half" as two unequal parts. Writing as a frontline primary mathematics teacher, this paper proposes two models: the "Error-Trace Bank (ETB)" and "Teacher-Orchestrated AI". Technology is responsible for preserving the process — rough drafts, verbal explanations, number-line clicks, figure partitions, and correction trajectories; the teacher is responsible for orchestration — deciding when to let AI generate a similar task, when to let a child explain to a peer, and when to deliberately withhold correction so the whole class can first see "this mistake is actually rather clever." Three lesson types — Grade 3 multiplicative meaning, Grade 4 fraction concepts, and Grade 5 data handling — show that digitalization is not about letting machines teach mathematics, but about letting machines make children's mathematical thinking visible, and then letting teachers turn what has been made visible into a lesson. The model is low-threshold, school-customizable, suitable for ordinary urban primary schools, and can be adapted for small-class and multicultural-student settings.

Keywords: educational digitalization; primary mathematics; Error-Trace Bank; teacher orchestration; AI digital textbook; process-oriented assessment