Evolution of Artificial Intelligence in Mathematics Education Research(2020-2025): From Student-Centered Personalization to Human–AI Pedagogical Intelligence in Mathematics Education

Artificial intelligence (AI) has become a major force in mathematics education, reshaping teaching, learning, assessment, and educational decision-making. Despite rapid growth in the literature, most reviews have focused on specific technologies or applications, providing limited insight into the field’s broader intellectual development. This study examines the evolution of AI in mathematics education research from 2020 to 2025 using SciMAT science-mapping analysis of 817 Scopus-indexed publications. Results reveal three developmental phases: learner-centered personalization (2020–2021), human–AI collaboration (2022–2023), and pedagogically integrated AI ecosystems (2024–2025). Early research emphasized students, intelligent tutoring systems, and chatbots; the second phase introduced mathematics teachers, adaptive learning systems, and large language models; and the most recent phase was characterized by teaching, ChatGPT, adaptive assessment, and AI-enabled pedagogical innovation. Thematic continuity combined with increasing specialization demonstrates that the field has evolved beyond technology adoption toward integrated pedagogical ecosystems. Based on these findings, the study proposes Human–AI Pedagogical Intelligence (HAPI), a framework that conceptualizes educational effectiveness as the interaction of teachers, learners, AI systems, and pedagogical practice. The study contributes a longitudinal understanding of the field’s development and provides directions for future research, policy, and responsible AI integration in mathematics education.