Data Set for the study "Evolving Constructs of Mathematical Problem Solving: A Topic Modeling Analysis of Research Trends and Cognitive Themes (1970–2025)"

This study explores the conceptual and thematic evolution of research on problem solving, problem posing, and problem-based instruction in mathematics education from 1970 to 2025. Using a large-scale text mining approach, 3,790 peer-reviewed articles indexed in Scopus and Web of Science were analyzed through Latent Dirichlet Allocation (LDA) topic modeling, trend analysis, and hierarchical clustering to identify dominant and emerging research directions. Six interrelated topics were identified: (1) cognitive development and mental processes in mathematical thinking, (2) instructional methods and pedagogical design, (3) conceptual understanding among diverse learners, (4) teachers’ roles and expertise in facilitating problem solving, (5) technology-enhanced and material-supported learning, and (6) psychological and metacognitive factors influencing performance. The findings reveal a steady expansion of research activity, particularly after 2000, with increasing attention to teacher cognition, student-centered pedagogies, and metacognitive regulation. While psychological and affective aspects of problem solving remain less represented, a marked rise in their prominence since 2020 suggests a renewed interest in internal cognitive and emotional mechanisms. Hierarchical clustering and multidimensional scaling analyses further indicate strong conceptual links between pedagogical and cognitive themes, reflecting a shift toward integrative, learner-centered approaches in mathematical thinking research. These results offer a comprehensive map of how the field has conceptualized problem solving over five decades, illuminating both saturated and emerging areas for future theoretical and empirical inquiry.