CBL is increasingly viewed not merely as a content-delivery modality but as a pedagogical ecosystem that supports the development of learners’ intellectual, metacognitive, and socio-communicative competencies (Masharova et al., 2020; Reddy et al., 2024; Suansokchuak and Piriyasurawong, 2025; UNESCO IITE, 2022). Such ecosystems provide on-demand access to distributed data storage and computing resources, as well as interactive platforms for knowledge exchange and collaborative work (Qashou et al., 2025). In doing so, they enable collaborative, self-regulated, and context-sensitive learning. The defining features of CBL scalability, interactivity, and reliability make it a particularly promising basis for resilient, adaptive, and more accessible educational models (Zhuoma, 2025).

In this study, CBL is defined as a pedagogically organized learning environment grounded in cloud infrastructure and internet-based platforms that provide sustained online access to learning materials, assignments, and interaction tools (Supplementary Tables A1, A2). This environment integrates digital services into a unified learning system in which students and teachers work with content and learning tasks through personal accounts, electronic gradebooks, interactive resources, and communication tools, thereby increasing the accessibility, personalization, relevance, and learnability of the educational process. CBL is operationalized as a sequence of learning actions (receiving an assignment, completing it, producing learning artefacts and, when necessary, editing them, engaging in discussion, receiving feedback, and revising the work accordingly). This iterative cycle is enabled by the cloud environment and simultaneously recorded as a verifiable digital trace of participants’ activity. The resulting traceability renders learning processes observable and therefore pedagogically manageable. Teachers can systematically scaffold students’ work, diagnose learning difficulties, provide targeted feedback, support iterative performance improvement, and foster the cognitive and collaborative skills required for learning (see Figure 1).

In this study, cognitive development is conceptualized as comprising two analytically distinct yet interrelated components: cognitive abilities (CA) and cognitive skills (CS). CA are closely intertwined with social–emotional competence (SEC) and provide a foundational basis for children’s school adjustment and their capacity to interact effectively with peers and adults (Clausén Gull et al., 2024). CA refer to relatively stable, domain-general mental capacities (e.g., executive functions, attention, working memory, and fluid reasoning) that constitute the cognitive infrastructure for learning. By contrast, CS denote trainable, task-specific intellectual operations targeted in this study’s Informatics outcomes (e.g., computational thinking, algorithmic problem solving, logical-analytical reasoning, and metacognitive strategies) that are enacted within instructional contexts. Self-regulated learning is treated in this study as a learning-process mechanism (mediator), rather than as a cognitive outcome skill (Zimmerman, 1990; Zimmerman, 2000). This distinction enables a theoretically grounded and empirically tractable analysis of lower-secondary students’ cognitive development. Abstract reasoning is among the most stable cognitive abilities and is typically only weakly amenable to short-term pedagogical influence. In the context of CBL, the learning environment provides students with more effective tools and strategies, but it does not imply changes in basic neurocognitive parameters, including information-processing speed and the overall efficiency of the cognitive system. Accordingly, CBL effects on CA are expected to manifest primarily through improvements in executive control (including attention and working memory), whereas gains in general intelligence are likely to remain limited or difficult to detect empirically.

Existing scholarship often discusses cloud technologies in education at a general level or frames them primarily as infrastructural enablers, without offering an integrated and empirically testable account of how CBL specifically reshapes learning processes and why such changes should translate into cognitive gains (Al-Samarraie and Saeed, 2018; Alamri, 2025; Noncheva, 2024; Yusuf, 2025). A substantial share of the evidence base comes from early childhood or higher-education settings, and findings are frequently generalized to lower-secondary schooling (Anupan and Chimmalee, 2024; Khasawneh, 2024; Qashou et al., 2025). This is problematic because adolescence is marked by rapid development of executive functions, metacognition, and higher-order reasoning, which may be particularly sensitive to instructional design features and feedback structures embedded in digital environments (Panadero, 2017; Sweller et al., 2019; Zimmerman, 2002).

Moreover, many studies operationalize CBL as access to an online platform rather than as a pedagogical configuration capable of restructuring cycles of learning actions (Baanqud et al., 2020; Olefirenko and Dobrunov, 2025; Reddy et al., 2024). Similarly, research on computational thinking and programming often evaluates isolated tools or instructional solutions, but rarely examines cloud-mediated mechanisms that could explain how learning activity and engagement change and why cognitive outcomes should consequently shift (Huang et al., 2025; Li et al., 2025; Sukkamart et al., 2025). A considerable portion of the literature relies on indirect indicators that are not equivalent to performance-based assessments of cognitive skills and abilities (Ahmed and El-Sabagh, 2025; Kharchenko et al., 2024; Kopcha and Sullivan, 2007; Orndorff, 2015). As a result, comparable evidence remains limited regarding whether CBL produces measurable improvements in target cognitive domains, rather than only shifts in attitudes and dispositions toward learning.

Robust findings that can disentangle CBL effects from selection processes, baseline between-group differences, and contextual confounding factors do not yet dominate the CBL literature (Chankaeva and Matsieva, 2023; Fadeeva, 2022; Gerasimova and Fadeeva, 2022; Pavlenko, 2023; Seidametova and Seitvelieva, 2011; Shekerbekova and Nesipkaliev, 2015). In addition, follow-up assessments are uncommon, leaving uncertainty as to whether observed improvements persist beyond the intervention or reflect short-term novelty and engagement effects (Matcha et al., 2020; Weydner-Volkmann and Bär, 2024). Even when positive effects are reported, relatively few studies model how institutional capacity, teacher readiness, and technology adoption barriers shape implementation fidelity and the depth of CBL integration and, by extension, variability in educational outcomes (de Reuver and Bouwman, 2015; Salloum et al., 2019; Scherer et al., 2019; Tsai and Chai, 2012). This gap is particularly salient for Kazakhstan, where digitalization priorities coexist with policy and implementation challenges documented in the literature (Adilet, 2023; Adilet, 2024; Baishan, 2022). Overall, the literature suggests that CBL is promising; however, adolescent-specific evidence remains limited (Al-Samarraie and Saeed, 2018; Baanqud et al., 2020; Matcha et al., 2020; Panadero, 2017; Sweller et al., 2019; Yusuf, 2025).

Therefore, the hypotheses of this study are formulated as follows:

We conceptualize CBL not only as a technological tool but also as an organizational and institutional innovation. First, infrastructure readiness reliable internet connectivity, adequate devices, and secure access to platforms provides the baseline for consistent cloud-supported instruction. Second, technology by itself does not guarantee pedagogically meaningful use: teachers’ digital and instructional competence determines whether cloud tools are integrated through cognitively demanding tasks, structured scaffolding, and formative feedback loops (Eickelmann and Vennemann, 2017). Third, sustained CBL implementation depends on financial and administrative viability (funding, maintenance, leadership commitment), which reduces fragmentation, supports continuity, and enables scaling over time (Dyusembekova et al., 2022). Finally, student and family readiness including home support, learners’ metacognitive maturity, and language/interface compatibility can either facilitate participation and equitable engagement or create barriers in cloud environments (Anupan and Chimmalee, 2024; Joshkun et al., 2024; Yeleussiz and Qanay, 2025).

Durable effects of CBL are feasible only when adequate infrastructural, organizational-regulatory, pedagogical, and family-home conditions are in place, ensuring regular and methodologically sound use of the digital environment in learning tasks and feedback practices. This multi-level perspective enables a more precise differentiation of each layer’s contribution within the educational ecosystem and supports the development of targeted practical strategies aligned with the degree of readiness at each level.

Grounded in Vygotsky’s sociocultural theory and the ZPD, cognitive development is expected to emerge through guided, socially mediated activity that is gradually internalized as self-regulated cognitive control (Vygotsky, 1980). In CBL-supported computer science instruction, cloud platforms can increase the visibility and continuity of instructional support (task sequencing, scaffolds, timely feedback, revision histories), which may strengthen students’ perceived teacher competence (PTC) and instructional guidance. Higher PTC, in turn, should foster self-regulated learning (SRL) and cognitive engagement (CE), enabling sustained effort on complex problem-solving tasks and supporting the development of higher-order cognitive skills (critical thinking, problem solving, logical reasoning, creativity, decision-making) (Aoonlamai and Kwangmuang, 2025; Zou et al., 2025). CS such as critical thinking, problem solving, logical reasoning, creativity, and decision-making are not merely individual traits but trainable forms of higher-order cognition that develop through structured practice, feedback, and scaffolding in meaningful tasks (Aoonlamai and Kwangmuang, 2025). Students who strengthen these skills typically demonstrate improvements in spatial visualization, logical inference, strategic problem solving, and constructive reasoning (Salloum et al., 2019). Accordingly, focusing on the development of such competencies through CBL aligns with international calls for evidence-based pedagogical interventions that improve learning outcomes and support durable academic competencies (Spirina et al., 2025). Assessing the impact of CBL on cognitive development therefore provides an empirical basis for evaluating the effectiveness of this approach in forming key competencies among lower-secondary students (Supplementary Tables A3, A4).

Self-regulated learning (SRL) is a core metacognitive process through which students intentionally manage cognitive resources and learning strategies to attain instructional goals. SRL comprises interrelated components of metacognition (planning, monitoring, and evaluation), motivation (sustaining effort and goal commitment), and strategic behavior (selecting, implementing, and adapting learning strategies). A substantial body of evidence links SRL to higher academic achievement and to learners’ capacity for lifelong learning, as students shift from passive information uptake to active regulation of attention, effort allocation, progress monitoring, and strategy adjustment when difficulties arise. Consequently, learning becomes more cognitively engaged, structured, and resilient to distraction, and the resulting knowledge is more meaningful and more readily transferable to novel learning situations (Belenkova, 2022; Berman and Bezmaternykh, 2023).

In many school settings, students report low motivation to learn computer science, difficulties with spatial visualization and logical thinking, and suboptimal academic performance challenges that reflect broader systemic constraints in digital education (Lei et al., 2025; Rivas et al., 2025). CBL can address these barriers by enabling interactive and socially embedded learning environments in which peer collaboration, scaffolded tasks, and formative feedback support the transition from externally guided performance to internally regulated cognitive autonomy. From a cognitive perspective, fosters students’ key metacognitive processes primarily through learning-process mechanisms by increasing PTC, which strengthens SRL and CE, thereby promoting gains in CS (Gazeykina and Kuvina, 2012; Gerasimova and Fadeeva, 2022). Conceptually, cloud computing is defined as on-demand network access to configurable computing resources delivered as services, emphasizing elasticity, scalability, and reliable shared access (Antipenko, 2016; Antipenko and Shkredova, 2019; Belenkova, 2022; Berman and Bezmaternykh, 2023; De Corte, 2019; Garkusha and Gorodova, 2023; Karpov et al., 2021; Karpovich and Koroleva, 2020).

From the perspective of Cognitive Load Theory, learning is optimized when instructional design reduces extraneous cognitive load and allows working-memory resources to be allocated to schema construction, conceptual integration, and transfer (Backsanskiy and Sorokina, 2024; Sweller, 2023; Sweller et al., 2019). Conventional computer science classrooms often impose high coordination demands due to fragmented resources, discontinuous feedback, and frequent switching between materials. In contrast, CBL environments consolidate resources, guidance, and assessment criteria within a single workspace, thereby reducing extraneous processing and supporting deeper semantic processing and knowledge transfer (Castro-Alonso et al., 2021; Hadie et al., 2021; Liu et al., 2022; Noetel et al., 2021; Rohde et al., 2023; Zhang et al., 2021). In conventional classroom settings, students often navigate fragmented resources (e.g., handouts, oral instructions, scattered files, and discontinuous feedback), increasing search, switching, and coordination demands classic sources of extraneous cognitive load. By contrast, CBL can consolidate materials and guidance within a single workspace, provide consistent navigation cues, and align tasks with resources and assessment criteria, thereby freeing cognitive capacity for schema construction and deeper processing. Moreover, when multimedia and interactive elements are designed in line with evidence-based instructional principles, they can clarify complex relationships, support representational integration, and scaffold schema development, thereby improving conceptual understanding and retention (Castro-Alonso et al., 2021; Gutiérrez Carreón, 2025; Hadie et al., 2021; Latifzadeh et al., 2021; Liu et al., 2022; Noetel et al., 2021; Rohde et al., 2023; Zhang et al., 2021; see Supplementary Table A6).

In many conventional classrooms, planning and monitoring depend largely on students’ internal tracking and intermittent teacher feedback, while learning traces, such as drafts, revisions, and feedback histories are often fragmented or short-lived. By contrast, CBL environments provide persistent records of learning activity and integrated scaffolds that support progress dashboards and visualization tools facilitate monitoring and pacing (Giorgashvili et al., 2024), and digital journals and formative assessment tools promote ongoing reflection and strategy adjustment (Dragomir and Niculescu, 2024). Digital scaffolds such as embedded prompts and peer-commenting features have been shown to increase metacognitive engagement (Sharma et al., 2024). Moreover, CBL can extend SRL beyond the individual to co-regulated learning. Collaborative platforms enable shared goal setting, mutual feedback, and coordinated strategy use in group projects, while personalized feedback mechanisms (including peer-comparison dashboards) may further strengthen metacognitive self-regulation and improve academic outcomes (Fleur et al., 2023). At the same time, these affordances require careful pedagogical and ethical design to support rather than constrain student autonomy (Weydner-Volkmann and Bär, 2024). Overall, when implemented thoughtfully, CBL shifts regulation from episodic, teacher-driven correction toward more continuous, learner-managed monitoring and adaptation, thereby strengthening SRL compared with traditional approaches. Therefore, we hypothesize a differentiated cognitive profile of effects:

CS is operationalized as a performance-based composite comprising computational thinking, algorithmic problem solving, logical-analytical reasoning, and metacognitive strategies. In contrast, improvements in CA are expected to be smaller and to reflect primarily enhanced executive control (e.g., attention and working memory), while changes in reasoning are expected to be limited.

CBL may act not only on students but also on teachers’ instructional capacity. This hypothesis is grounded in the activity theory (Leontiev, 1978) and constructivist paradigms (Bruner, 1990; Vygotsky, 1980), which consider learning as a purposeful, socioculturally mediated, and cognitively rich process. The pedagogical structure of classroom activities formed by a teacher determines whether the technology become an instrument for cognitive growth or remain a mere channel for information delivery. Constructivist principles, such as co-construction of knowledge, reflection, and transfer, require teachers to design tasks that actively engage students in modeling, discussion, and peer evaluation. Ahmed and El-Sabagh (2025) show that a gamified cloud-based LMS can yield significantly higher engagement and achievement than a conventional LMS, highlighting that CBL effects depend on pedagogical adaptation, not platform access alone. Similarly, Zhang et al. (2025) emphasize that CBL yields cognitive benefits only when embedded in constructivist-oriented designs, including peer review, knowledge visualization, and reflective practices. Educational outcomes are determined by the effectiveness of the pedagogical methodology of technology use rather than by access to it. As Shrestha (2025) further demonstrates, students’ cognitive engagement intensifies solely when tools are employed within pedagogically sound instructional scenarios. Thus, technological novelty or scalability alone does not guarantee educational effectiveness; rather, it is the quality of didactic integration that determines meaningful learning outcomes. The impact of CBL on cognitive development is mediated through learning objectives, task structure and complexity, the logic of digital interaction among participants, and the availability of formative reflective tools.

Teachers’ readiness to integrate CBL is grounded in the Technology Acceptance Model (TAM), which posits that perceived usefulness and perceived ease of use are key determinants of technology adoption (Davis, 1989). In educational contexts, these perceptions are activated during lesson planning, selection of digital tools, and the design of learning activities. Crucially, teachers’ pedagogical reflection on the compatibility of new solutions with their existing instructional approaches plays a decisive role in shaping adoption decisions. Empirical studies confirm this link. Scherer et al. (2019) found that teachers’ perceptions of usefulness and ease of use are directly associated with their willingness to employ digital tools to support students’ cognitive development. Similarly, the success of digital integration varies across national contexts, depending on teachers’ subjective beliefs and their professional autonomy. Moreover, teachers’ motivation to adopt CBL is significantly stronger when digital tools are perceived as complementing, rather than replacing, established pedagogical practices. This is particularly relevant in personalized learning environments, where the goal is to foster students’ cognitive autonomy, self-regulation, and sustained engagement, rather than automate instruction.

The research model (Figure 2) specifies a multilevel implementation–process–outcome framework linking institutional and contextual readiness conditions and teacher-level instructional capacity to learning-process mechanisms and students’ cognitive outcomes under CBL.

Accordingly, the study is guided by the following research questions:

RQ1. To what extent do institutional and contextual readiness factors predict the level of CBL integration in schools, and does school context (e.g., urban vs. rural setting) moderate these relationships?

Participants were lower-secondary students (Grades 8–9) enrolled in compulsory Informatics courses during the 2024/2025 academic year. The sampling frame comprised approximately 4,581 students from 18 public schools located in both urban and rural regions (Table 1). Using stratified cluster sampling, we selected 66 intact classes from these schools to ensure proportional representation by school context (urban/rural), geographic area, and school-level socioeconomic indicators. The final analytic sample included 1,650 students across the 66 classes (Table 2).

Classes were allocated to the experimental (CBL) and control conditions in a 1:1 ratio, using stratified randomization at the class level to balance school context and baseline academic indicators (Tables 1, 2). Baseline equivalence between conditions was examined for key demographic characteristics and prior academic indicators and was confirmed prior to the intervention (Table 3).

To minimize contamination and between-group influence, we implemented both procedural and technical controls. First, intervention teachers received CBL-specific onboarding and instructional materials, whereas control teachers did not receive access to the CBL platforms or training during the study period. Second, cloud workspaces and LMS course areas were created specifically for the experimental classes and were restricted through class-specific accounts and access permissions such as whitelisted class rosters and password-protected environments. Third, experimental and control classes were scheduled separately where feasible, and students were instructed not to share credentials or cloud materials across classes. Together, these measures reduced the likelihood of cross-condition exposure and helped preserve the internal validity of the classroom-level comparison.

The study employed an explanatory mixed-methods design combining standardized cognitive tasks, validated questionnaires, and semi-structured interviews and focus groups. This design enabled the concurrent assessment of cognitive-development outcomes, hypothesized learning mechanisms, and institutional-pedagogical conditions shaping CBL implementation (see Table 5).

Students’ cognitive development was assessed at four time points: pre-intervention (T0), mid-intervention (T1), immediately post-intervention (T2), and 12 weeks after completion (T3) to examine the durability of effects (see Tables 5, 6). All tasks were administered individually in classroom settings using standardized instructions and fixed time limits. Teachers delivering the lessons did not administer the tests, reducing the risk of assessment bias.

CS were operationalized across four performance-based domains:

Raw scores from all cognitive tasks were converted into z-scores, with standardization conducted separately for each measurement wave (T0-T3), ensuring appropriate comparability across assessment occasions (Table 6).

Throughout the intervention, systematic observations were conducted in both experimental and control classrooms. A structured protocol captured behavioral activity, technology-use patterns, and instructional quality, including feedback characteristics and pedagogical support (Table 8). The full data-collection protocol, including detailed information on instruments and coding systems, is provided in Supplementary Tables A5–A10.

To assess hypothesized learning mechanisms and implementation conditions, we used validated questionnaires adapted for an adolescent sample (Tables 7, 9–11). All instruments employed Likert-type scales (5–7 points). Internal consistency was evaluated using Cronbach’s α (Tables 7, 9, 10).

Qualitative data were collected to deepen the analysis of mechanisms identified in the quantitative strand and to interpret the institutional conditions of CBL implementation. Semi-structured interviews were conducted with teachers, and student focus groups (6–8 participants each) were held. All sessions were audio-recorded. Qualitative materials were used to explain differences in the enactment of CBL tools and observed instructional practices (see Tables 8, 15).

Quantitative findings provided effect-size estimates for the intervention and tests of the hypothesized mechanisms, whereas qualitative data were used to contextualize and explain the observed implementation patterns. Quantitative analyses followed a pre-specified sequence. Distributional normality was assessed using the Shapiro–Wilk test; when needed, additional model assumptions were evaluated at the residual level. Baseline equivalence between the experimental and control groups was examined using independent-samples t-tests and standardized mean differences (SMD) (see Table 3). Within-group changes from T0 to T3 were analyzed using paired t-tests, with primary attention to posttest (T2) and follow-up (T3) comparisons where applicable (see Table 16 for effects at T2 and durability at T3).

The primary between-group intervention effect was tested using ANCOVA, with posttest outcomes as dependent variables and baseline scores entered as covariates. Practical significance was evaluated via partial η² (Table 17) and Hedges’ g (for standardized between-group differences; Table 16). Associations among key variables were examined using Pearson correlation coefficients. All statistical analyses were conducted in SPSS.

Hypotheses derived from the conceptual model were tested using structural equation modeling (SEM) in AMOS. Several hypotheses were specified through multiple structurally linked relationships (e.g., H2 was modeled as PTC predicting SRL and PTC predicting CE; H3 as SRL predicting CA and CE predicting CA; and H4 as SRL predicting CS and CE predicting CS) (Table 11). The model’s explanatory power was assessed using squared multiple correlations (SMC/R²) for each endogenous variable; these values are reported in Table 14 and align with the coefficients of multiple determination summarized in Table 13. In addition, an auxiliary regression specification including Teacher Readiness (TR) was estimated to quantify the pedagogical contribution beyond student mediators; unstandardized coefficients are reported in Table 18.

Interview and focus-group data were analyzed using thematic analysis. Two researchers independently coded transcripts using a theoretically informed codebook aligned with the study’s conceptual framework; discrepancies were resolved through discussion to ensure coding consistency. Themes related to changes in engagement, learning regulation, and technology use were used to interpret the quantitative findings and were illustrated with representative quotations. Overall, integrating quantitative and qualitative procedures enhanced both the statistical rigor of effect estimation and the interpretability of results for classroom practice.

The intervention was implemented over twelve consecutive weeks, with two Informatics lessons per week, ensuring sustained exposure to CBL components while aligning with the national curriculum. This duration supports meaningful skill development and allows for integration into routine classroom practice. Prior to implementation, a preparatory phase ensured technical infrastructure readiness across all participating schools, standardized platform configurations, and teacher training for the experimental group. All procedures followed international guidelines for intervention reporting, ensuring transparency, replicability, and methodological rigor throughout the study.

Structural equation modeling (SEM) was conducted in AMOS (v. 20.0) to estimate direct, indirect, and total effects of key variables; results are presented in Table 12.

Several factors were consistently associated with stronger gains in the experimental group. In particular, robust digital infrastructure, higher teacher competence, and stronger family support predicted improvements in intellectual abilities, in line with H1a, H1b, and H1d. Teacher readiness and effectiveness also showed a meaningful contribution to student outcomes (H5). At the student level, metacognitive awareness, baseline digital literacy, and task persistence strengthened the intervention’s impact, supporting H2.

By contrast, some variables were not related to cognitive outcomes in this model, including platform switching frequency, collaboration mode (individual vs. group), student age, and total time spent online. Taken together, the results suggest that cognitive growth depends less on the sheer amount of technology use and more on how the tools are embedded in instruction and supported by the surrounding context.

The convergence of evidence across pre/post cognitive trajectories (Supplementary Table A12), predictors of intellectual skills regarding SRL and CE interactions (Supplementary Table A15), and academic performance (Table 5) confirms the systemic and sustained nature of CBL impact. The effects are sustainable; key mediating mechanisms include the structured cognitive architecture of the course design, intensification of self-regulation, and maintenance of deep cognitive engagement.

The cloud-based pedagogical intervention exerted a statistically significant and educationally meaningful impact on students’ cognitive development, comparable to the results of studies that use elements of structural modeling to assess the direct and indirect influences of the learning environment The post-test effect size (Hedges’ g = 0.21) aligns with benchmarks for a moderate effect in educational research, indicating practical relevance for large-scale schooling environments. Given that the study was conducted under naturalistic conditions without altering curricular content or instructional load, this estimate is conservative though pedagogically robust (Supplementary Tables A13, A14).

Extended comparative analyses revealed the strongest effects on computational thinking (g = 0.31) and algorithmic problem-solving (g = 0.25), both with 95% confidence intervals excluding zero, which indicates the reliability of the effect (Table 16). Slightly lower but statistically significant gains were observed in logical-analytical reasoning (g = 0.20) and metacognitive strategies (g = 0.24). This pattern complies with the logic of the intervention design: cloud-based tasks were structured as a sequence of visually supported tasks, primarily activating self-regulatory and problem-solving operations, which subsequently enhanced higher-order cognitive outcomes.

To assess sustainability of the observed differences a delayed post-test (T3) was administered 12 weeks after intervention completion. Results confirmed the persistence of effects, with an even stronger aggregate effect (Hedges’ g = 0.28, p = 0.025), indicating that experimental-group advantages in cognitive capabilities were maintained over time. This suggests the development of stable digital work habits and self-regulatory routines in students’ learning.

These findings align with the presented factorial model of digital education: pedagogical and organizational enablers (teacher preparedness use Information and Communication Technologies, quality of instructional design, infrastructure reliability, and family support) were key to fostering computational and regulatory competencies. Crucially, the recorded statistically significant effect stems not only from technological density of the educational process alone, but also from thoughtfully designed digital practice, integrated into the structure of the lesson and supported by self-regulation mechanisms.

Notably, influence of external factors, which are not entirely controllable, should be considered. However, the consistency of effect sizes (Hedges’ g values) across immediate and delayed assessments, alongside robust regression coefficients, supports the methodological validity and scalability of this digital pedagogical intervention in diverse school settings for replication in school practice and for inclusion in expanded models for assessing the quality of digitalization of education.

To examine the study’s theoretical logic, we estimated a structural equation model (SEM) specifying a sequential mechanism through which CBL influences students’ cognitive development via PTC, SRL, and CE. Model fit indices indicated an excellent fit to the data: χ²(36) = 49.73, p > 0.05; RMR = 0.041; GFI = 0.948; AGFI = 0.921; CFI = 0.975; TLI = 0.963; RMSEA = 0.045. Final path estimates corresponding to the tested hypotheses are reported in Table 11, and the explained variance (SMC/R²) for the key endogenous constructs is presented in Table 14.

The measurement component of the model included six reflective indicators each for CA and CS (5-point Likert scale) and was validated on the pooled sample comprising both the experimental and control groups. All scales demonstrated high reliability and convergent validity. Specifically, Cronbach’s α ranged from 0.76 to 0.92, exceeding the 0.70 threshold; composite reliability (CR) was above 0.70; and AVE values (0.58–0.82) met the criterion of explaining more than 50% of the variance in indicators through the latent construct (Table 9). Standardized factor loadings were high (all > 0.75, p < 0.001), indicating strong indicator-construct relations and no weak items in the questionnaire structure (Table 10). Descriptive statistics (means and standard deviations) showed sufficient response variability to support valid comparisons between the experimental and control groups.

CBL significantly increased PTC, and higher PTC in turn strengthened SRL and CE. SRL and CE then jointly predicted cognitive outcomes (Table 11). The model explained 66.4% of the variance in CA and 64.7% of the variance in CS (Tables 11, 13, 14, 17, 20), indicating substantial explanatory power for a naturalistic educational design. Significant indirect effects from PTC to both CA and CS further indicate that CBL operates predominantly through mediated pathways via students’ engagement and self-regulation rather than through mere exposure to technology (Tables 11, 20).

Beyond the student-centered CBL pathway where CBL is expected to improve cognitive outcomes through higher PTC, stronger SRL, and greater CE we examined Teacher Readiness (TR) as an additional mechanism. Three facets were assessed: readiness to design cloud-based tasks (TR1), the frequency and quality of cognitively oriented digital practices (TR2), and teachers’ self-efficacy in the CBL environment (TR3). Overall, TR was judged to be sufficient, consistent with Hypothesis H5 (Table 8).

To clarify how the teacher factor contributes to the intervention logic, we estimated supplementary regression models. The results show that CBL is positively and significantly associated with higher TR, and that higher TR, in turn, is associated with better student cognitive outcomes (Table 18). These findings therefore support a dual account of CBL effects. One pathway operates through students’ experiences via increased PTC, SRL, and CE (Tables 11, 14) while a second pathway operates through teachers via strengthened readiness and the transformation of instructional practices (Tables 8, 18). Taken together, the evidence supports all five hypotheses and aligns with the proposed multi-level conceptual model. In this sense, CBL should be understood not as a technology added to instruction, but as a pedagogically structured learning environment that improves the quality of classroom processes and teaching practices and, through these improvements, contributes to gains in students’ cognitive outcomes.

First, the study design supports only a limited causal interpretation. Randomization was implemented at the classroom level within schools, whereas school selection relied on convenience sampling. Consequently, the influence of unobserved school-level factors cannot be fully ruled out. Nevertheless, the risk of systematic baseline differences was mitigated by the confirmed equivalence of groups at the start of the experiment. Second, the generalizability of the findings is constrained. Although the sample included 18 schools from multiple regions with an equal urban–rural distribution, it is not statistically representative of all school types in Kazakhstan, particularly schools in socially vulnerable and severely resource-constrained areas. Moreover, even under the stated inclusion criteria, substantial heterogeneity in infrastructure and implementation conditions was observed, especially in rural schools. This limits the transferability of effect estimates and highlights the need for further research using broader and more diverse samples. Third, several key process variables were measured using self-reports, and self-report bias cannot be fully eliminated. Fourth, while infrastructural differences and implementation parameters were described (Tables 4, 15), the detailed assessment of pedagogical implementation quality was limited. As a result, some portion of effect variability may reflect differences in instructional enactment that were not fully captured by the measurement framework. Fifth, conclusions regarding the teacher-related component should be interpreted cautiously. Indicators of teacher readiness and self-efficacy showed statistically significant between-group differences (Table 8), and the model links teacher readiness to cognitive outcomes (Table 18). However, the number of teachers included in the analysis was limited, which may reduce external validity. It is also important to clarify the extent to which practice indicators were derived from self-reports versus classroom observations. Future studies should therefore involve larger teacher samples and independent protocols for assessing pedagogical practice. Finally, the observation period and assessment of sustainability were limited. The intervention and primary posttest covered one semester, supporting inference about short-term effects, whereas longitudinal designs with longer follow-up and explicit transfer assessment are required.

Future research should prioritize longitudinal designs to examine whether the effects of CBL on cognitive development persist beyond short-term interventions, including follow-up assessments at 6–12 months, and to test whether observed gains transfer to other school subjects and task types. At the design level, multicenter and cross-cultural studies are needed to compare how digital transformation unfolds in education systems with different levels of infrastructure, teacher readiness, and institutional maturity. Such work would help specify boundary conditions of CBL effectiveness and improve the portability of conclusions. Methodologically, it is advisable to expand the scope of randomized and quasi-experimental research by using larger and more heterogeneous samples and by including underrepresented groups, such as rural schools, students with diverse language backgrounds, and socioeconomically vulnerable populations. This would strengthen causal inference regarding links between digital tools, instructional design, and cognitive outcomes, while also improving generalizability. A critical direction remains the development and validation of psychometrically robust instruments for assessing cognitive outcomes. Promising approaches include combining standardized performance-based measures with platform log data, interaction patterns, task-completion sequences, and behavioral indicators, which can reduce reliance on self-report scales and lower the risk of systematic measurement bias. Finally, dedicated attention is required to analyze the digital divide and its implications for vulnerable student groups, including those facing social, linguistic, or cognitive barriers. Such evidence is essential for designing equitable, evidence-based, and sustainable digitalization strategies aimed at reducing disparities between schools and student groups.

Within Kazakhstan’s ongoing educational transformation, the present findings underscore the need to shift from an infrastructure-oriented approach toward a systemic digitalization model in which teacher readiness and instructional design are central. Such a model would secure not only access to digital tools, but also the sustained development of universal cognitive competencies required for success in a knowledge-based economy and a digitally mediated society. Teacher readiness is heterogeneous and depends both on school-level internet connectivity and cloud infrastructure, as well as on teachers’ overall digital competence. In practice, teachers tend to respond positively to professional development when it is practice-oriented, embedded in students’ authentic learning activities, and when implementation does not increase administrative burden but instead yields clear instructional benefits. Accordingly, to scale CBL across primary and lower-secondary education and reduce disparities between urban and rural schools, several measures are recommended: ensuring high-speed and secure cloud infrastructure in more than 98% of schools; delivering practice-oriented professional development to at least 85% of informatics and mathematics teachers with a focus on cloud-task design, scaffolding, and assessment; establishing a national repository of validated, curriculum-aligned digital tasks with templates and rubrics; and implementing a national monitoring and analytics system with KPIs that capture not only coverage, but also implementation quality and students’ cognitive and academic outcomes. These steps should be coordinated under a unified national strategy that links technology access with pedagogical capacity, not as isolated inputs but as interdependent components of a sustainable digital learning system.