Designing an AI-Supported Formative Assessment Model for Pre-Service Mathematics Teacher Self-Study in Vietnam

Abstract

Amid Vietnam’s accelerated digital transformation, teacher education is undergoing a pivotal shift—demanding innovative pedagogical and assessment strategies to foster self-directed learning (SDL) competencies. This need is particularly acute in Mathematics, a discipline that requires advanced logical reasoning, autonomy, and metacognitive engagement. In response, this study proposes a theoretically grounded and empirically validated formative assessment (FA) model designed to enhance SDL among pre-service primary Mathematics teachers. The model integrates both internal factors (e.g., motivation, metacognitive skills, and self-assessment competences) and external elements (e.g., technological tools, instructional support, and learning environments). Data were collected from 438 pre-service Mathematics teachers across multiple universities and analyzed using structural equation modeling (SEM). The findings identify two primary predictors of effective SDL: learners’ perceptions and attitudes toward FA and their intrinsic motivation and persistence. These factors significantly impact learning outcomes and are further amplified when supported by AI-driven tools and digital learning platforms. The study highlights the mediating role of technological and pedagogical scaffolds in strengthening SDL and emphasizes the critical integration of formative feedback, metacognitive strategies, and adaptive instructional design. Despite encouraging results, several challenges remain, including limited learner autonomy, inconsistent technology adoption, and reliance on instructor-led practices. This study contributes to a deeper understanding of how AI-enhanced FA can promote learner autonomy, critical thinking, and sustainable learning behaviors in Mathematics education. It offers actionable insights for policymakers and educators in designing resilient, context-sensitive teacher training frameworks aligned with the demands of the digital era. Future research should focus on piloting the model across diverse educational settings and leveraging big data and emerging technologies to improve predictive validity and pedagogical impact.