Human-centred AI pedagogy in primary mathematics: a theoretical framework for agency and sense-making
Artificial intelligence (AI) can expand the resources available for primary mathematics teaching, but assisted task completion is not the same as mathematical learning. This article develops a Human-Centred AI Pedagogy Framework through a synthesis of sociocultural theory and humans-with-media, human-centred AI and hybrid intelligence, social cognitive theory, TPACK and its AI extensions, and Ubuntu-informed relational ethics. The framework connects pedagogical intentionality, teacher agency, learner agency, mathematical sense-making, and ethical and relational responsibility. Its proposed explanatory contribution concerns the destination of mathematical decisions in a teacher–child–AI relationship. When teachers reduce direct guidance, decisions may be delegated to children or taken over by the system; this paper describes the latter possibility as decision drift. Agency handover, therefore, means increasing children's supported responsibility for strategy, representation, justification, and help-seeking, while teachers retain safeguarding and instructional accountability. The dimensions are differentiated as prerequisites, enabling conditions, mediating processes, outcomes and ethical boundaries. Two hypothetical lessons, on Grade 3 place value and Grade 5 growing patterns, illustrate the framework and compare AI-supported and non-AI alternatives. South African multilingualism, unequal infrastructure and teacher workload specify important implementation conditions. Five propositions identify expected relationships, observable indicators and competing explanations, providing a basis for classroom observation and comparative studies. Evidence from older learners motivates the analysis but does not establish effects in primary classrooms. The framework's explanatory and predictive validity remains to be tested; its value is a more precise account of what AI-mediated mathematical activity should enable, protect and make open to investigation.
Authors
- Zanele Dibane (ORCID: https://orcid.org/0000-0002-9121-7744)
Institutions
- University of South Africa (ZA)
Publication Details
- Journal
- Frontiers in Education
- Published
- 2026-11-19
- DOI
- https://doi.org/10.3389/feduc.2026.2043389
- Primary Topic
- Artificial Intelligence in Education
- Type
- article
- Field-Weighted Citation Impact
- 0.00