Introduction: Mathematical skills play a significant role in advancing educational outcomes. These skills enhance the ways individuals perform numerous tasks in life (Onoshakpokaiye, 2023; Parviainen, 2019). Mathematical literacy ranks among the fundamental components of the 21st century skills, equipping students with real-world problem-solving methods and helping them meeting social demands (Rizki & Priatna, 2019).

Given the importance of mathematical competence, Trends in International Mathematics and Science Study (TIMSS) examines contextual factors associated with mathematics achievement. Among several factors, teacher instructional clarity has been reported as a basis for better mathematics scores (Fadiji & Reddy, 2023), while student digital skills have gained increasing attention for their valuable impact on contemporary, technology-oriented learning environments (Bergdahl et al., 2020; Youssef et al., 2022).

Kosovo was ranked among the lower-performing education systems in TIMSS 2023 (Von Davier et al., 2024). However, the existing literature on this topic remains still limited, indicating a gap in exploring the Kosovar context and factors associated with it. In this regard, this study aimed to investigate the predictive effect of teacher instructional clarity and student digital skills on mathematics achievement using TIMSS 2023 data for Kosovo, examining to what extent these two factors predict students’ mathematical achievement.

Methods: The study employed a quantitative (Slater & Hasson, 2024), correlational approach, consistent with its objective to examine predictive relationships of the variables involved (Tan, 2014). Using the stratified-cluster, random sampling method, the sample consisted of 4,609 fourth-graders who participated in TIMSS 2023, with data collected through the Mathematics Achievement Test and Student Questionnaire. The data were publicly available for research purposes and did not involve identifiable information (Fishbein et al., 2025); therefore, no ethical or formal approval was required for this study. Upon completion of all recoding procedures, the data were further processed using SPSS, RStudio, and the IEA IDB Analyzer.

Results and Discussion: The findings from descriptive analysis revealed that students reported high levels of teacher instructional clarity (x̄ = 3.80, SD = 0.42) and advanced digital skills (x̄ = 3.15, SD = 0.63). Both constructs were positively correlated with student mathematics achievement in TIMSS 2023, such that higher teacher instructional clarity (r = .22, p < .001) and more advanced digital skills (r = .16, p < .001) were associated with higher mathematics achievement scores.

Further, multiple regression analysis conducted in the IDB Analyzer confirmed that teacher instructional clarity significantly predicted mathematics performance (t = 7.40, p < .001), as did student digital skills (t = 6.88, p < .001). It is fundamental to note that both predictors combined together explained a meaningful portion of the variance in mathematics achievement (R2 = .07, Adj. R2 = .07), emphasizing the importance of teacher clarity and digital competence in supporting mathematical skills.

These findings suggest that students have a greater chance to understand and apply mathematical concepts when their teachers present clearly (Fadiji & Reddy, 2023). These teachers are likely to promote better understanding (Pinter et al., 2017), which further leads to greater mathematics achievement. This result aligns with previous research (e.g., Berger et al., 2023; Mowahed, 2023), underscoring that developing clear instructional skill plays a pivotal role in advancing mathematics scores.

In addition, the positive effect that student digital skills have on mathematics achievement (Pagani et al., 2016) indicates that these students have mastered their digital competence and are able to use digital technology for learning and problem-solving (Sari et al., 2020), which leads to increasing mathematics performance. These findings are in line with previous studies (for example, Busnawir et al., 2023; Li et al., 2025; Joshi et al., 2025; Standhope & Celestine, 2025; Yagan, 2021) that emphasize the need for initiatives that promote innovative skills in learning environments.

Educational significance of the research: This study makes meaningful contributions to the educational practice, policy, and future research by improving mathematical instruction and supporting digital literacy development. In this regard, the findings of the study provide useful insights into strengthening student-teacher interactions, advancing digital competence, enhancing teacher professional development, guiding teacher support systems and digitalization of school resources, informing policy development and curriculum design, extending the existing literature, and fostering future research in the area. Taken together, all these factors are closely linked to improving students’ mathematics performance in the future.

Introduction

Despite widespread use of online learning environments, evidence on their benefits for skill acquisition remains inconsistent. A critical explanation for this discrepancy is insufficient practice, as only about 5% of students sustain recommended weekly engagement with online learning environments for mathematics learning environments (e.g., Holt, 2024). This raises a fundamental question for the field of technology-enhanced science education: how can we better capture the behavioral dynamics determining whether students persist long enough, and with sufficient effort, to benefit from practice in online learning environments? Given the availability of fine-grained log data that capture students decisions to persist, seek challenges, or quit, research on online learning is uniquely positioned to contribute to solving the issue of low engagement by identifying which behavioral facets matter most for engaging with and persisting to use respective online learning environments (e.g., Andrade et al., 2026). In this study, we conceptualize engagement and persistence as longitudinal choices to start, continue, quit, or retry problems and problem sets in an online environment for learning mathematics. We were interested in how these behaviors correlate with one another and predict future performance on heldout data, yielding interpretable indicators and dimensions for effort-aware adaptivity.

Methods

Leveraging large-scale log data from the digital learning environment bettermarks (N = 2,129 students; >780,000 problem-set records), we estimate latent dimensions of these behaviors and their factor structure before validating our model’s predictive power for future engagement and performance. In particular, we employed latent-trait modeling based on student and item parameters, for instance, where college student ability is estimated based on their grade achievement after adjusting for the types of courses they enrolled in (Baucks et al.,2024). We use these methods to produce novel, content–adjusted, and log-based student trait measures of engagement and persistence using Rasch-style generalized linear mixed models (GLMMs). We then apply a novel trait-modeling approach to estimate latent dimensions of student engagement and persistence from logs of start, continue, quit, and retry decisions during learning mathematics in an online learning environment. Each opportunity to continue (after a problem, hint, or error) is treated as a decision point, and the probability of persisting is modeled as a function of recent behavioral signals. Finally, we examine the resulting structure of latent dimensions to evaluate its stability and predictive validity.

Results and Discussion

Results provided a novel content–adjusted, log-based conceptualization of engagement and persistence estimated based on learner propensities, which we then use in a cross-validated factor analysis to understand the latent dimensionality of persistence and engagement. We observed that all three identified factors—Completion Propensity, and Error-Free Completion, and Activity—are reliable across time and prospectively predictive on held-out data. In particular, Completion Propensity predicts retrying and the selection of harder content despite lower accuracy; Error-Free Completion predicts higher first-attempt accuracy alongside a preference for easier content; and Activity predicts retry behavior while being associated with lower accuracy. Importantly, the three-factor model explained substantially more variance in future learner behavior than a unidimensional model (25% vs. 3%, respectively).

Thereby, this paper reframes engagement and persistence as sequences of micro-choices to start, continue, quit, and retry, rather than as coarse usage totals. From cross-validated factor analysis on content-adjusted Rasch-style person parameters, we established three reliable, prospectively predictive dimensions. This parsimonious structure explained substantially more variance in future behavior than a single score, why completion propensity aligns with retrying and appropriate challenge, and why an error-free orientation often coincides with preference for easier material. The resulting indicators enable effort-aware adaptivity that surfaces who persists, how they persist, and where to intervene, complementing knowledge models with calibrated effort profiles.

Educational Significance of the research

These profiles can be used to complement existing cognitive tutoring (e.g., by adjusting difficulty) or to serve insights into teachers orchestrating use of the learning environments (e.g., for improving assessment and content selection). Beyond methodological value, these indicators afford practical leverage for effort-aware adaptivity by distinguishing students who continue despite difficulty from those who optimize for error-free performance or gravitate towards easier content.

1. Introduction

In response to the digital transformation, Austria introduced the mandatory school subject Basic Digital Education (BDE) in the school year 2022/2023 (BGBl. II Nr. 267/2022) in lower secondary education. However, the sudden implementation of BDE has left many teachers struggling. In turn, teaching practices within BDE are highly influenced by teachers’ personal knowledge, interests and values. Here, special emphasis should be placed on teachers’ attitudes towards students’ belonging. While belonging (students’ sense of relatedness and inclusion) is linked to motivation and achievement (Goodenow, 1993; Ryan & Deci, 2000), little is known about how teachers’ attitudes shape belonging in digital education. Thus, this study examines teachers' attitudes towards students’ belonging in digital education.

2. Methods

In this research lecture I will focus on eight semi-structured interviews with BDE teachers conducted in April and May 2024. Using a combined analysis method – theme identification with thematic analysis (TA) and attitudes/typology reconstruction with documentary method (DM) – this study examined teachers’ values placed on academic achievement, school belonging, power imbalances, relationships and normative beliefs. Participants provided informed consent, all data was anonymised, and the study was approved by the ethics committee.

3. Results and Discussion

The DM suggests two main types of teachers in digital education: Type I teachers who prioritise traditional teaching practices and type II teachers who prioritise student-centred teaching practices. Across interviews, four belonging-related attitudes were identified: (1) dichotomisation of academic achievement and school belonging, (2) power imbalances within teacher-student-relationship, (3) practices that foster student-student-relationship and (4) an unmarked norm privileging non-disabled, German-speaking students. Type I teachers more often prioritised academic achievement and maintained stronger power asymmetries within teaching practices, which possibly roots in the competitiveness of schools (Heinz, 2023). Type II teachers tended to value peer relationships and collaborative, student-centred practices. Thus, we find the withdrawal of strong hierarchies and punishments and a development towards strong teacher-student-relationships in digital education within type II. To support this movement, structural conditions need to be adapted, i. e. more time for digital education and subject specific teacher training for BDE teachers.

In line with Yuval-Davis (2006) teachers’ attitudes towards students’ belonging in digital education are related to several norms that are rooted in political ideologies. Especially disabled students or students with a migration background seem to be disadvantaged in relation to digital education. They are either marked as needy or excluded from class. Notably, two interviews challenged the unmarked norm across types. Further, it is important to note that individual teachers do not actively exclude marked students from their classes, but structural mechanisms make the inclusion of said students difficult. The analysis further suggests that teacher-ascribed categories (i.e. reading difficulties) solidify students’ identities in ways that restrict their capacity to be seen otherwise, with implications for belonging. This resonates with Spies (2019) on subjectification through addressing.

4. Educational Significance of the Research

The educational significance of this study lies in its interdisciplinary framing of belonging as a key lever in BDE. From motivational psychology, belonging is a basic need that drives engagement, persistence, and learning (Ryan & Deci, 2000). Sociologically, the analysis of addressing practices and the unmarked norm shows how categorisations (i. e. reading difficulties) shape subjectivities and access to recognition, with direct consequences for inclusion and belonging (Spies, 2019; Yuval-Davis, 2006). This implies that task design, collaboration structures, and assessment must cultivate social connectedness. Otherwise, BDE may reproduce digital divides: higher disengagement, selective participation, and widening achievement gaps.