Who is Doing the Thinking? Dialogic Conditions in AI-Mediated Primary Mathematics

Source abstract

Policy and procurement are moving faster than evidence in the rush to place LLM tutors in primary classrooms. Debate has largely asked whether AI tutoring raises attainment, yet this research argues that the more important question is whether it sustains or closes down the dialogic conditions through which children’s mathematical reasoning develops. Drawing on dialogic learning theory, mathematical epistemology and a postdigital critique, it reframes AI tutoring as a pedagogical setup in which the cognitive work of mathematical reasoning is distributed across child, teacher and AI. The research’s central contribution is a two-dimensional analytical framework that distinguishes the dialogic quality of an interaction from its mathematical quality. The framework displays a particular risk of fluent AI dialogue, where interactions look conversational while the reasoning is supplied to, rather than developed by, the child. A proposed three-phase study to validate and apply the framework in upper Key Stage 2 mathematics is outlined. The aim is a practical lens for evaluating AI tutors by the learning relationships that they produce, and not only by the answers they provide.