"I can, so I teach with AI": Modelling the interplay between self-efficacy, mathematics-related beliefs and TPACK-M

Abstract

This study investigated, through structural equation modelling, the relationships among pre-service elementary teachers' beliefs about mathematics teaching and learning, their self-efficacy regarding artificial intelligence (AI) usage and their mathematics-specific technological pedagogical content knowledge (TPACK-M). The participants comprised pre-service elementary teachers selected from three public universities in Türkiye. Data were collected using the Mathematics-Related Beliefs Scale, AI Self-Efficacy Scale and Mathematics Technological Pedagogical Content Knowledge Scale. The findings revealed that AI self-efficacy has a direct positive effect on TPACK-M. Constructivist beliefs positively influence TPACK-M development through both direct and indirect pathways. While traditional mathematics beliefs had no significant direct effect, they negatively influenced TPACK-M indirectly through AI self-efficacy. Consequently, the proposed model indicates that AI self-efficacy serves as an essential mediating variable in developing TPACK-M, constructivist beliefs significantly enhance this development, whereas traditional beliefs exhibit a restrictive role.

Implications for practice or policy:

• Teacher education programme designers should ground mathematics lessons in constructivist pedagogy because constructivist beliefs directly strengthen TPACK-M.
• Primary education faculty should design low-risk AI tasks because AI self-efficacy is the strongest direct predictor of TPACK-M.
• Policymakers in Türkiye's Council of Higher Education and Ministry of National Education should align primary teacher education standards with constructivist principles because traditional beliefs lower AI self-efficacy, which in turn constrains TPACK-M development.