Deep Learning Research Trends in Mathematics Learning: A Scoping Review And Bibliometric Analysis (ScoRBA)

This study aims to map the research landscape of deep learning in mathematics education using bibliometric mapping based on Scopus indexed data from 2011 to 2025. Using scientific mapping methods via VOSviewer and R Bibliometrix, a total of 201 publitions meeting the inclusion criteria were analyzed. To address construct validity, the study explicitly differentiates between deep learning as a computational artificial intelligence (AI) methodology and deep learning as a pedagogical concept (deep vs. surface learning), analyzing their respective representations across clusters. The findings demonstrate: (1) a rapid growth trajectory in publications, peaking at 45 articles in 2025; (2) the top contributing journals are Education Sciences, Eurasia Journal of Mathematics, Science and Technology Education, Frontiers in Psychology, International Journal of Mathematical Education in Science and Technology (5 articles each), and Educational Studies in Mathematics (4 articles); (3) Y.F. Zakariya is the most prolific author (4 articles); (4) Beijing Normal University (China) and Universitetet i Agder (Norway) are the most productive institutional affiliations; (5) the United States leads global production (61 articles), followed by China (28 articles) and Australia (16 articles), with Indonesia ranking fourth (11 articles); and (6) three primary thematic clusters were delineated: the Red Cluster (AI computational models and e learning applications), the Green Cluster (pedagogical constructs of mathematical deep learning and cognition), and the Blue Cluster (STEM/STEAM integration and instructional design). Emerging frontiers highlight the integration of generative AI at the elementary education level and cross disciplinary STEAM frameworks, offering promising avenues for future research.