Referee Report
Title: SmartFlex Learning Ecosystem: Integrating AI and Learning Analytics in Vocational Mathematics Education
Summary of the Article
This manuscript presents the development and evaluation of a SmartFlex Learning Ecosystem, an integrated educational model that combines artificial intelligence (AI), learning analytics, adaptive learning, and flexible learning strategies within vocational mathematics education. Using a Design-Based Research (DBR) framework, the authors report the implementation of the SmartFlex ecosystem with 172 vocational students over a 12-week period. Quantitative analyses using Partial Least Squares Structural Equation Modelling (PLS-SEM) indicate significant positive relationships among SmartFlex, learning engagement, adaptive learning, mathematical competence, and vocational skill readiness. The authors conclude that SmartFlex provides a novel framework that can improve vocational mathematics education through AI-enhanced personalised learning and data-informed instructional support.
The topic is timely, relevant, and important given the increasing adoption of AI and learning analytics in education. The manuscript is well structured and demonstrates considerable effort in integrating contemporary educational technologies. Nevertheless, the scientific contribution is weakened by substantial conceptual, theoretical, methodological, and interpretative shortcomings. The study currently presents a promising implementation rather than a convincing theoretical or methodological advancement.

Overall Evaluation
The manuscript addresses an important educational problem and has potential practical value. However, in its current form it does not fully justify its claims of novelty, theoretical advancement, or methodological innovation. While the statistical analyses are generally competently performed, they are not matched by equivalent rigor in the conceptual framing, research design, or interpretation of findings. The manuscript would benefit from substantial revision before it can be considered scientifically robust.

Major Strengths
The study addresses a contemporary and relevant issue in vocational mathematics education by integrating AI and learning analytics into a single instructional environment. The manuscript is logically organised, the objectives are clearly stated, and the statistical reporting follows accepted PLS-SEM conventions. The authors have also made commendable efforts to improve transparency by depositing underlying datasets and supplementary materials in an open repository. The practical implications for educators and policymakers are significant and may inform future implementations of AI-supported vocational learning systems.

Major Concerns
1. The manuscript lacks a genuine theoretical foundation (Must Address)
The most significant weakness of the manuscript is its lack of explicit theoretical grounding. Although the literature review discusses artificial intelligence, learning analytics, adaptive learning, HyFlex learning, and vocational education, these concepts are not anchored within established educational or learning theories. Instead, hypotheses are justified primarily through previous empirical studies rather than theoretical propositions.
Consequently, the conceptual framework lacks explanatory depth. The manuscript repeatedly claims to extend HyFlex learning and to contribute theoretically to educational technology; however, no educational theory is explicitly adopted, operationalised, or revisited in interpreting the findings.
Several well-established theories would have strengthened the study considerably, including Self-Regulated Learning Theory, Constructivism, Self-Determination Theory, Social Cognitive Theory, Cognitive Load Theory, Activity Theory, or Expectancy-Value Theory. Without such theoretical grounding, the proposed SmartFlex model appears largely as an integration of existing technologies rather than a theoretically informed educational innovation.
Recommendation: Develop a clear theoretical framework that informs the conceptual model, hypotheses, data interpretation, and claimed contribution.

2. The novelty of the SmartFlex Learning Ecosystem is overstated (Must Address)
Throughout the manuscript, SmartFlex is described as a novel educational ecosystem. However, the proposed model largely combines existing concepts that have already been extensively investigated, including AI-driven personalisation, adaptive learning, learning analytics, competency-based education, and flexible learning.
The manuscript does not convincingly demonstrate how SmartFlex differs conceptually from existing smart learning ecosystem models beyond integrating these components under a new label. The novelty, therefore, appears incremental rather than transformative.
Recommendation: Clearly distinguish SmartFlex from existing smart learning ecosystem frameworks through systematic comparison and explicitly state what constitutes its original contribution.

3. The Design-Based Research methodology is insufficiently implemented (Must Address)
The manuscript identifies the study as Design-Based Research (DBR), yet the reported methodology resembles a conventional implementation study followed by structural equation modelling. Authentic DBR requires iterative design cycles, continuous refinement, documented design principles, and theory generation.
Although four DBR phases are described conceptually, the manuscript provides little evidence of iterative redesign, modifications between cycles, design decisions, or knowledge generated through the design process. The study, therefore, does not fully satisfy accepted standards for DBR.
Recommendation: Either provide detailed documentation of the iterative DBR process or reconsider whether DBR is the most appropriate methodological label.

4. The claimed mixed-methods design is not adequately reported (Must Address)
The methods section describes a sequential mixed-methods design involving questionnaires, learning analytics, interviews, and classroom observations. However, the Results section reports almost exclusively quantitative findings. No qualitative themes, interview quotations, coding framework, or integrated mixed-methods interpretation are presented.
Consequently, the manuscript does not demonstrate that qualitative evidence informed the reported conclusions.
Recommendation: Present the qualitative findings, explain the thematic analysis procedures, and demonstrate how qualitative and quantitative findings were integrated. Alternatively, revise the methodological description if the qualitative component played only a minor role.

5. Research questions are not fully answered (Must Address)
The manuscript formulates eight research questions, including questions concerning AI literacy, teacher beliefs, ethics, accessibility, educational data mining, and AI-driven assessment. However, the empirical analyses address only the relationships represented within the PLS-SEM model. Research Questions 5–8 are neither analysed nor discussed.
This inconsistency weakens the coherence of the study and leaves several stated objectives unanswered.
Recommendation: Either address these research questions empirically or remove them to ensure consistency between aims, methods, results, and conclusions.

6. Methodological transparency is insufficient for replication (Must Address)
Although the manuscript describes the general research procedures, numerous methodological details remain absent.
The manuscript does not sufficiently explain:

• the SmartFlex platform architecture;
• AI algorithms and personalisation mechanisms;
• instructional activities during the intervention;
• development and validation of measurement instruments;
• questionnaire items;
• qualitative protocols;
• analytical workflow;
• intervention materials.

Consequently, independent replication would be difficult despite the availability of datasets.
Recommendation: Substantially expand the methods section and, where possible, provide additional implementation materials.

7. Statistical interpretation is stronger than warranted (Should Address)
The statistical analyses themselves are generally appropriate, and the reported measurement model demonstrates acceptable reliability and validity. However, the interpretation occasionally exceeds what the evidence supports.
The study relies on purposive sampling and a single implementation, yet the discussion frequently implies causal relationships and broad educational effectiveness. Furthermore, all ten hypotheses are supported with moderate-to-strong effects and no unexpected findings, but potential biases such as common method variance, sampling limitations, and alternative explanations are not critically discussed.
Recommendation: Moderate causal language, report additional diagnostic analyses where appropriate, and discuss alternative interpretations of the findings.

8. The discussion lacks critical interpretation (Should Address)
The discussion primarily restates the statistical findings and links them to previous studies. It provides limited explanation of why the observed relationships occurred, under what conditions they may differ, or how they relate to educational theory.
A stronger discussion should critically interpret the findings, consider competing explanations, acknowledge contradictory evidence, and explicitly discuss the implications for theory.

9. The literature review requires greater synthesis (Should Address)
The literature review is comprehensive but largely descriptive. Paragraphs typically summarise recent studies individually rather than synthesise themes, identify inconsistencies, or critically evaluate existing knowledge.
The review also relies heavily on publications from 2024–2026, with relatively limited engagement with seminal literature in AI in education, learning analytics, mathematics education, and instructional design.
Recommendation: Incorporate foundational literature and strengthen the critical synthesis of previous research.

10. Conclusions overstate the evidence (Should Address)
The conclusions generally reflect the reported statistical relationships but extend beyond what the evidence supports.
The manuscript claims theoretical advancement, empirical validation of SmartFlex, and broad educational applicability. However, these claims are based on a single implementation in a single vocational context and are insufficiently supported by the study design.
The conclusions should therefore be moderated to reflect the exploratory and context-specific nature of the findings.

Minor Comments
The manuscript would benefit from reducing repetition, particularly regarding phrases such as "AI-driven personalisation," "adaptive learning," and "learning engagement." Several sections repeat similar arguments without adding substantive insight.
Some methodological terminology also requires clarification, particularly the distinction between SmartFlex as a framework, ecosystem, instructional model, or technological platform.
Finally, greater consistency among the conceptual framework, research questions, hypotheses, and conclusions would improve the manuscript's overall coherence.

Responses to the Review Questions
1. Is the work clearly and accurately presented and does it cite the current literature?
Response: Partly
The manuscript is generally well organised and cites a large body of recent literature. However, the presentation contains important inconsistencies, particularly between the stated research questions and reported analyses. The literature review is current but largely descriptive and insufficiently grounded in foundational scholarship.

2. Is the study design appropriate and is the work technically sound?
Response: Partly
The statistical procedures are generally appropriate, but the study design is weakened by inconsistencies between the claimed Design-Based Research approach, the reported mixed-methods design, and the empirical analyses actually presented.

3. Are sufficient details of methods and analysis provided to allow replication?
Response: Partly
Although the general procedures are described, insufficient information is provided regarding the intervention, instruments, AI implementation, and iterative DBR process to permit full replication.

4. Is the statistical analysis and its interpretation appropriate?
Response: Partly
The PLS-SEM analysis is competently conducted and appropriately reported. However, interpretation of the findings occasionally exceeds the evidence, and important diagnostic analyses and limitations are insufficiently discussed.

5. Are all source data available?
Response: Partly
The deposited datasets improve transparency, but they are insufficient on their own to reproduce the intervention, implementation process, and complete analytical workflow.

6. Are the conclusions adequately supported by the results?
Response: Partly
The conclusions are generally supported by the reported statistical associations but overstate the novelty, theoretical contribution, causal implications, and generalisability of the findings.

Issues That Must Be Addressed Before the Article Can Be Considered Scientifically Sound
The following issues are fundamental and should be addressed before publication:

1. Establish a clear theoretical framework underpinning the conceptual model and hypotheses.
2. Justify the novelty of the SmartFlex Learning Ecosystem through comparison with existing models.
3. Demonstrate authentic Design-Based Research implementation or revise the methodological description.
4. Report and integrate the qualitative component of the claimed mixed-methods design.
5. Ensure all research questions are answered or revise the study aims accordingly.
6. Provide substantially greater methodological detail to enable replication.
7. Moderate claims regarding theoretical contribution, causal inference, and generalisability.
8. Strengthen the discussion through critical interpretation rather than descriptive restatement.
9. Improve the literature review through greater synthesis and inclusion of foundational scholarship.
10. Align the conclusions strictly with the evidence presented.

Final Recommendation
This manuscript has the potential to make a useful applied contribution to the field of AI-supported vocational mathematics education. However, its current scientific contribution is limited by insufficient theoretical grounding, methodological inconsistencies, and overstated claims of novelty and validation. The statistical analysis is generally acceptable, but strong empirical modelling alone cannot compensate for weaknesses in conceptualisation and research design.
Accordingly, I recommend Major Revision. If the authors successfully address the fundamental concerns identified above, the manuscript could become a valuable contribution to the literature. In its present form, however, it does not yet meet the standards expected of a rigorous, theory-driven educational research article suitable for publication in a high-quality international journal.