Hwa Kyung Kim , Minho Song , Inyong Choi

DOI:10.63311/mathedu.26.6521 JANT Vol.65(No.2) 223-246, 2026

---

This study aims to explore the feasibility of automated scoring of mathematical constructed-response items using the fine-tuning of a large language model (LLM). To this end, the performance of fine-tuning was analyzed by varying the amount of training data, and implications were drawn accordingly. First, constructed-response tasks in three distinct areas of mathematics were developed, student responses were collected, and expert scoring was conducted. The results of automated scoring performed by a publicly available base LLM―provided with items and scoring rubrics―were then compared with those obtained after fine-tuning. The results indicated that scoring performance improved as the amount of training data increased, with a substantial improvement observed when at least 30 training samples per score category were secured. Furthermore, the degree of performance improvement varied depending on the characteristics of the task. Based on these findings, implications for the educational application of fine-tuning in mathematics education were discussed.

Jiyoung Kim

DOI:10.63311/mathedu.26.6522 JANT Vol.65(No.2) 247-265, 2026

---

This study analyzes how the newly introduced achievement standard “[6수04-06] predicting possibilities and making judgments based on data” is implemented in elementary mathematics textbooks aligned with the 2022 revised Korean mathematics curriculum. A census analysis was conducted on all nine sets of authorized elementary mathematics textbooks, comprising seven that developed the standard into independent lessons and two that integrated it into adjacent lessons. The analytical framework combined the epistemic status of data (data-as-answer vs. data-as-evidence) and levels of reasoning (L1-L4). The results indicate that in most of the nine textbooks, data primarily functioned as data-as-answer, serving as information to finalize a judgment. Reasoning paths consistently converged at the L3a level, where students progressed from reading data (L1) and identifying relationships (L2) to providing a single, deterministic judgment. Although publishers employed diverse design strategies ― such as pattern induction through multiple trials, reverse reasoning using small samples, and large-scale trials across multiple tools ― the final expected responses were invariably deterministic. A partial exception was identified in one publisher’s discussion activity, which presented conflicting student interpretations of the same data and accepted multiple responses as valid, thereby corresponding to L3b and partially incorporating elements of L4-level reflection on uncertainty. These findings suggest that the structural convergence toward deterministic judgment reflects not individual authorial choices but systemic tendencies rooted in the ambiguity of curriculum guidelines, the nature of the textbook authorization system, and implicit assumptions about developmental appropriateness. This study provides foundational data for efforts to more substantively realize the core essence of data-based statistical reasoning ― uncertainty and tentativeness ― in elementary statistics education, and offers direction for the refinement of curriculum guidelines and textbook development.

Sehyung Lee , Soomin Park , Gaeun Ji , Sunghwan Moon , Seung-jo Jung

DOI:10.63311/mathedu.26.6523 JANT Vol.65(No.2) 267-290, 2026

---

This study quantitatively analyzes the curriculum structure of mathematics education departments in Korean colleges of education and examines how these institutional arrangements relate to in-service mathematics teachers’ practical needs, based on curriculum data and teacher perceptions. The curriculum analysis indicates a clear disciplinary content-oriented sequence: mathematics content courses account for 68.9% of total credits and are arranged from the early years, whereas mathematics education courses become more prominent in the later years. Notably, a discrepancy was found between nominal course titles and their actual functions, as several courses categorized as pedagogy-oriented incorporate substantial content components, making it difficult to infer their instructional function from titles alone. Survey results identify major barriers in classroom practice, including limited understanding of teaching and learning methods, assessment methods and components, and curriculum interpretation. Specific qualitative accounts from the open-ended responses further suggest that these challenges are not merely a lack of isolated skills but stem from the complex nature of professional practice, which requires the organic integration of various elements. These elements include calibrating instructional goals and difficulty levels across wide ranges of student ability levels, designing performance assessments while ensuring fairness, translating standards into justifiable classroom decisions, and articulating the value of mathematics to students. Overall, the findings suggest that increasing disciplinary content alone does not guarantee teacher professional competence; teacher education curricula should be qualitatively restructured to provide repeated and scaffolded opportunities across all academic years for transforming disciplinary knowledge into classroom design, assessment, and decisions responsive to learners.

Jinam Hwang , Jin-kon Hong

DOI:10.63311/mathedu.26.6524 JANT Vol.65(No.2) 291-307, 2026

---

This study aims to examine how terms are introduced and adapted in the geometry and measurement domain of elementary school mathematics textbooks, with a focus on identifying the distance between scholarly knowledge and knowledge to be taught. To this end, elementary mathematics textbooks based on the 3rd national curriculum and the 2022 revised curriculum were selected for comparison. Drawing on the theory of didactic transposition, this study analyzes the sequence and structure of term introduction, as well as the patterns of term adaptation in both sets of textbooks. The results indicate that the 2022 revised textbooks tend to introduce whole geometric figures prior to their components, present terms in a way that reflects an intended progression from general to specific concepts, and rarely redefine previously introduced terms. In addition, terms that had been explicitly defined in the earlier textbooks were found to be adapted through the omission of definitions, substitution, relocation, or deletion, thereby increasing the distance from scholarly knowledge. These findings suggest that textbooks are products of knowledge recontextualized within institutional settings and that the introduction and adaptation of terms are closely related to the structure of concepts and their epistemological justification. This study further highlights the need for a balanced approach to the introduction and adaptation of terms in the development of future textbooks and curriculum, taking into account both the structure of disciplinary knowledge and learners’ understanding.

Yun Hwa Noh , Hee-jeong Kim

DOI:10.63311/mathedu.26.6525 JANT Vol.65(No.2) 309-335, 2026

---

The integration of AI-based digital tools into classroom instruction is transforming the nature of instructional practices required of teachers in mathematics classrooms. This study aimed to analyze the types of instrumental orchestration and teacher roles observed in high school mathematics classes utilizing AI-based digital tools. Six lessons taught by an experienced high school mathematics teacher with extensive experience using AI-based digital tools were observed, and data from lesson recordings, teacher tablet screen captures, and post-lesson interviews were analyzed using directed content analysis. The findings revealed four existing orchestration types―technicalsupport, explain-the-screen, guide-and-explain, and spot-and-show―alongside three newly emerging types arising from the distinctive characteristics of AI-based tools: dashboard-monitoring, AI-correcting, and AI-mediating. In addition, the teacher’s roles varied across orchestration types and included pedagogical roles, AI-technological roles, content-related roles, ethical roles, and the role of knowledge transmitter. Notably, content-related and ethical roles were particularly prominent when the teacher intervened in AI-generated feedback. By empirically illustrating how teacher orchestration is reconfigured in AI-based instructional environments, this study extends the theoretical framework of instrumental orchestration to AI-supported digital learning contexts and provides implications for understanding and supporting mathematics teachers’ roles and professional development in the age of AI.

Jihyun Hwang , Yeonseok Ka , Jinho Kim

DOI:10.63311/mathedu.26.6526 JANT Vol.65(No.2) 337-353, 2026

---

This study examined the longitudinal trajectories of changes in elementary preservice teachers’ beliefs about mathematics teaching and learning over a 15-week constructivist mathematics education course. A total of 80 preservice teachers were assessed four times (Weeks 1, 6, 9, and 12) on teacher-centered beliefs (TC) and studentcentered beliefs (SC). Longitudinal cluster analysis was conducted using the SFclust algorithm. The results indicated that belief changes did not converge toward a single average trajectory; rather, they were classified into four distinct clusters characterized by different developmental patterns. Variable weighting analysis showed that TC (0.974) had a substantially higher weight than SC (0.226), suggesting that differentiation among clusters was primarily driven by the degree to which existing teacher-centered beliefs weakened, rather than by the simple accumulation of student-centered beliefs. Notably, in all but one cluster, a sharp decline in TC was observed between Weeks 1 and 6, indicating that belief change tended to concentrate during the early phase of the course. Although SC scores were already relatively high at Week 1 across all clusters, additional increases in SC were observed only in clusters where TC showed substantial decline. These findings suggest that belief change may not be a process of merely adopting new beliefs, but rather one that involves the critical restructuring of pre-existing beliefs. By capturing heterogeneous developmental trajectories through longitudinal cluster analysis, this study provides a methodological framework for understanding preservice teachers’ belief transitions more precisely within the Korean educational context. The findings further offer implications for designing differentiated teacher education programs that take preservice teachers’ initial belief states into account.

Sung-yong Kwon

DOI:10.63311/mathedu.26.6527 JANT Vol.65(No.2) 355-378, 2026

---

This study explored how pre-service elementary teachers’ mathematical beliefs are configured into different profiles and examined how these profiles relate to mathematics preference and perceived mathematical ability. It also analyzed open-ended responses to interpret the experiential and affective characteristics of a group in which relatively high beliefs about mathematics learning did not align with positive mathematics preference. Using an explanatory sequential mixed methods design, k-means cluster analysis (k=3) was conducted with data from 227 third-year pre-service teachers at G University of Education. Three exploratory belief profiles were identified: a relatively traditional-oriented profile, a relatively constructivist-oriented profile, and an asymmetric nature-learning belief profile. The constructivist-oriented profile showed the highest proportion of mathematics preference, whereas the asymmetric nature-learning belief profile showed the lowest proportion; however, the overall group difference approached but did not reach statistical significance at the 0.05 level. Logistic regression analysis indicated that mathematics preference was not significantly associated with belief scores alone, whereas perceived mathematical ability showed a closer association with mathematics preference under the present measurement conditions. Qualitative analysis of open-ended responses showed that the high-learning-belief-low-preference group used language related to achievement and enjoyment less frequently, while more often expressing themes related to evaluation pressure, negative past learning experiences, and experiential explanations of asynchrony, with cognitive burden and anxiety also appearing as related tendencies. These findings suggest that pre-service teachers’ mathematics preference may be better understood by considering not only mathematical beliefs, but also perceived ability and the affective traces of accumulated learning experiences. This study proposes belief-preference asynchrony as an exploratory framework for interpreting the possibility that mathematical beliefs and mathematics preference do not necessarily develop or align in the same way. More specifically, the empirical focus of this study lies in learning-belief-preference asynchrony―cases in which comparatively high beliefs about mathematics learning do not necessarily accompany positive mathematics preference. However, mathematics preference in this study was operationalized as a limited indicator captured by a single dichotomous item (“I like mathematics”) and therefore should not be interpreted as a comprehensive measure of mathematics attitude.

Jieun Yuk , Hwayoung Lee , Hokyoung Ko

DOI:10.63311/mathedu.26.6528 JANT Vol.65(No.2) 379-401, 2026

---

This study examined the structural relationships among affective characteristics in mathematics learning using structural equation modeling, with growth mindset as the starting point. Specifically, it analyzed the pathways through which value perception, interest, and self-confidence influence challenge and perseverance as well as mathematics anxiety via the mediating role of learning motivation, and verified differences across school levels through multi-group analysis. Data were collected from 31,022 students across 1,153 schools―drawn from the population of elementary, middle, and high schools under 17 provincial and metropolitan offices of education nationwide―using stratified two-stage cluster sampling stratified by school level, region, and area size. A 25- item instrument comprising seven subfactors measured on a five-point Likert scale was employed. The results showed that growth mindset had significant positive effects on value perception (β=0.710), interest (β=0.562), and self-confidence (β=0.508), and that these variables, with learning motivation as the key mediator, strengthened challenge and perseverance while reducing mathematics anxiety. In the multi-group analysis, the effect of growth mindset was significant across all school levels; however, the paths from affective variables to learning motivation were frequently non-significant at the middle school level, confirming structural differences across school levels. In particular, the negative effect of learning motivation on mathematics anxiety was strongest at the high school level. These findings suggest that growth mindset education should be integrated with the internalization of value perception, and that school-level differentiated motivational support systems are needed.