The research team followed the PRISMA reporting protocol, the international standard for systematic reviews, to identify 163 Scopus-indexed articles on blended learning in programming education published over the 24-year window. Rather than reading each paper in isolation, the authors treated the corpus as a network to be mined. Using the Bibliometrix package in RStudio alongside Python-based preprocessing of the text, they analyzed 477 author keywords, traced co-citation patterns to reveal which works scholars actually build upon, and applied clustering algorithms to group the literature into coherent thematic families. This science-mapping approach, well established in fields such as management and information science, allows researchers to see the intellectual skeleton of a discipline rather than just its surface vocabulary.

Four primary domains emerged from the analysis. The first covers learning models themselves, most prominently the flipped classroom, in which students encounter instructional content such as video lectures before class and devote contact time to problem-solving, along with broader hybrid configurations that mix synchronous and asynchronous delivery. The second domain concerns technology platforms, including massive open online courses and learning management systems built on Moodle, which supply the infrastructure that makes blending practical at scale. The third domain captures persistent pedagogical challenges: sustaining student engagement, scaling instruction to very large cohorts, and designing assessment that works across physical and digital environments. The fourth domain documents interdisciplinary links, connecting programming education to STEM integration, computational thinking, and vocational training.

The temporal story the map tells is striking. Early studies in the 2000s focused on foundational questions of whether blended formats could deliver programming content at all, with researchers experimenting with e-learning platforms, Web 2.0 tools, and multi-sensory approaches. The 2010s brought an explosion of flipped classroom experiments, clicker-based interactivity studies, and investigations of learner characteristics that predict success in blended settings. Notably, some of the most influential work addressed equity and scale, including designs for teaching programming to very large classes in Sub-Saharan Africa, where blended approaches were framed as a response to resource constraints rather than a technological luxury. The COVID-19 pandemic then acted as an accelerant, pushing institutions worldwide into emergency remote teaching and generating a wave of studies on what happens when blending becomes mandatory rather than optional.

To probe the current frontier, the authors conducted a close review of two contrasting samples: the 25 most-cited papers in the corpus and the 25 most recent ones. The comparison exposed a sharp divergence between established theory and emerging practice. The highly cited literature rests on mature pedagogical frameworks, carefully designed interventions, and empirical measures of learning outcomes, engagement, and cognitive load. The most recent papers, by contrast, are dominated by large language models such as ChatGPT, which now feature prominently as tools for generating code explanations, supporting debugging, and providing personalized assistance to novice programmers.

That dominance, the study argues, is running well ahead of the evidence. While LLM-based tools are the subject of intense scholarly and popular enthusiasm, the authors found that empirical research rigorously testing their pedagogical impact in programming education remains limited. Much of the recent literature consists of perceptions, feasibility reports, and speculative frameworks rather than controlled studies measuring whether AI assistance actually improves students’ ability to understand and write code. This matters because programming education has long wrestled with high failure rates and fragile conceptual understanding, and interventions that look promising on the surface can mask shallow learning or encourage over-reliance. Related systematic reviews cited in the study raise concerns that heavy dependence on AI dialogue systems may erode the very critical-thinking and problem-solving skills that programming courses exist to develop.

The tension is not unique to programming instruction, but it is particularly acute there. Learning to program requires learners to construct mental models of abstract machine behavior, to debug their own reasoning as much as their code, and to persist through productive struggle. If a chatbot supplies polished solutions instantly, the formative difficulty that drives learning may simply evaporate. Conversely, well-designed AI scaffolding could offer the kind of individualized, around-the-clock support that blended formats have always promised but rarely delivered at scale. The study’s authors are careful not to declare a winner; their point is that the field currently lacks the empirical foundation to make that judgment, and that the gap between hype and evidence is widening as publication volume surges.

The knowledge map also highlights structural imbalances in the research itself. Studies cluster around introductory courses and higher education, with comparatively less attention to K-12 contexts, professional training, and learners in low-connectivity regions. Assessment remains a weak link: automated grading systems and portfolio-based evaluation appear in the literature, but coherent models for assessing learning across blended environments are still immature. The co-citation analysis further reveals that many recent papers cite a narrow band of foundational works, suggesting that new research may be under-engaged with the field’s accumulated theoretical base even as it chases novel technologies.

For educators and institutions, the study offers a practical consolidation. It confirms that flipped and hybrid models, when thoughtfully designed around active learning, have a substantial evidence trail behind them, and that platforms such as Moodle-based systems and MOOCs are now mature components of the programming education ecosystem. At the same time, it urges caution: the integration of generative AI into blended programming courses should be treated as a hypothesis to be tested, not a settled best practice. The authors call for rigorous empirical evaluation, including controlled comparisons and longitudinal studies, to guide evidence-based adoption of AI tools.

In an era when every syllabus is being rewritten around chatbots, this bibliometric map arrives as both a retrospective and a corrective. It shows how far blended learning in programming education has traveled, from early experiments with online content delivery to today’s AI-infused classrooms, and it identifies precisely where the evidence runs out. The field’s next chapter, the study suggests, should be written not by the fastest adopters but by the most careful measurers, ensuring that the tools transforming how students learn to code are held to the same empirical standard as the pedagogies they are replacing.

Subject of Research: Bibliometric mapping of blended learning research trends in programming education from 2000 to 2023

Article Title: From flip to click: a global bibliometric analysis of blended learning trends in programming education (2000–2023)

Article References: From flip to click: a global bibliometric analysis of blended learning trends in programming education (2000–2023). (n.d.). https://doi.org/10.1007/s43545-026-01760-7

Image Credits: AI Generated

DOI: 10.1007/s43545-026-01760-7

Keywords: blended learning, programming education, bibliometric analysis, flipped classroom, large language models, ChatGPT, MOOCs, Moodle, educational technology, computer science education, AI in education, PRISMA

The research, conducted by a team at Universitas Negeri Semarang and published in Discover Education, recruited participants through e-Guru.id, an Indonesian online professional learning network that offers webinars, memberships, and specialized training. All respondents had completed at least one MOOC related to professional development. The sample was predominantly female, with 1,665 women and 866 men, and professionally seasoned: nearly 78 percent were aged 35 or older, and almost 57 percent had at least 15 years of classroom experience. Notably, despite that deep experience, most were newcomers to MOOC learning, with 82.3 percent reporting zero to two years of MOOC use. The authors caution that because recruitment relied on a digitally connected professional network, the sample should not be treated as nationally representative of all Indonesian teachers.

Methodologically, the study is a cross-sectional survey designed to describe and explore rather than to test causal claims. Each of three conceptual domains was measured with five items on a four-point Likert scale, a format chosen to push participants toward a directional response rather than a neutral midpoint. The three domains were Perceived MOOC Effectiveness, covering instructional value, relevance, and professional usefulness; Perceived Professional Competence and Confidence, covering motivation and the perceived ability to apply new knowledge; and Perceived Opportunities and Challenges, covering scheduling flexibility, collaboration, effort, institutional support, and workload. Before any group comparisons, the team validated the instrument with confirmatory factor analysis using IBM SPSS Amos, and the three-factor model showed excellent fit, with a Comparative Fit Index of 0.972, a Tucker-Lewis Index of 0.967, and a Root Mean Square Error of Approximation of 0.057.

The statistical machinery behind the subgroup comparisons deserves attention because it reflects a growing standard of rigor in educational research. Rather than relying on classical analysis of variance, the team used Welch’s independent-samples t-tests and Welch’s one-way ANOVAs, which do not assume equal variances and are better suited to the substantially unequal group sizes in the sample. Because 15 item-level comparisons were run for each demographic variable, the researchers applied Benjamini-Hochberg false discovery rate corrections within each family of tests, and they followed significant omnibus results with Games-Howell post hoc tests. Effect sizes were reported as Cohen’s d for gender comparisons and as eta squared and omega squared for the ANOVAs. Of 75 item-level comparisons across five grouping variables, only three survived correction, a result the authors interpret as evidence of greater similarity than difference across teacher groups.

The descriptive findings tell a coherent story about what teachers value. Within the effectiveness domain, item means ranged from 3.42 to 3.64 on the four-point scale, with the highest rating going to the statement that MOOCs provide valuable resources and materials for improving teaching practice. Close behind were items on the contribution of MOOCs to understanding pedagogical strategies and instructional design, the relevance of content to classroom teaching, and usefulness for professional growth. The lowest item in the domain, though still positive, concerned perceived benefits for student learning, suggesting that teachers were more confident identifying immediate professional gains for themselves than attributing changes in student outcomes to their course participation. The authors stress that student outcomes were not measured independently, so this item reflects perception rather than demonstrated achievement gains.

The competence and confidence results reveal a subtle but important gap. Teachers rated their motivation to teach and share knowledge after MOOC participation highest, at a mean of 3.56, followed closely by perceived improvement in teaching skills. But confidence in applying MOOC-acquired knowledge in practice and in integrating technology into classroom teaching scored slightly lower, at 3.46 and 3.45 respectively. The authors argue this distinction separates professional learning acquisition from professional learning transfer: teachers may absorb knowledge and feel motivated, yet still struggle to translate that knowledge into technology-supported classroom practice. Because the data are self-reported and cross-sectional, the findings capture perceptions of capability rather than independently verified changes in instructional performance.

The opportunities-and-challenges domain produced the most differentiated pattern and the study’s most striking single result. Teachers rated opportunities to collaborate with colleagues across Indonesia highly, at a mean of 3.50, and valued the flexibility of MOOCs amid busy schedules at 3.45. Yet the item measuring difficulty completing MOOCs because of workload or other commitments scored an identical 3.45, and the effort required to keep up with course content scored 3.39. Flexibility, in other words, did not eliminate the demands of participation; it reorganized them. The lowest mean across the entire survey went to institutional support, at just 3.03, with a large standard deviation of 0.92 indicating wide variation in teachers’ experiences of organizational backing. The authors suggest that many participants may not have experienced MOOC learning as embedded within school-level professional development structures.

The subgroup analysis dismantles several common assumptions about who benefits from online professional learning. After false discovery rate correction, age, educational attainment, and teaching experience showed no robust differences on any of the 15 items, challenging narratives that frame digital learning through generational or qualification-based divides. Only two small gender differences survived correction: male teachers reported slightly higher perceived scheduling flexibility and slightly stronger institutional support, with Cohen’s d values of just -0.12 and -0.13. A nominal gender gap in technology-integration confidence vanished after adjustment. The one substantive pattern involved duration of MOOC use: teachers with more than six years of experience rated content relevance significantly higher than newer users, though the omnibus effect was small, explaining less than 1 percent of variance, and the longest-duration group contained only 74 participants. The authors note that causality cannot be inferred; sustained participants may find courses more relevant, or prolonged engagement may sharpen their ability to identify relevant content.

What emerges is a socio-technical picture in which MOOCs are not standalone solutions but components of broader professional learning ecosystems. The authors argue that access to online courses does not automatically become meaningful professional growth, and that knowledge acquired through MOOCs requires opportunities for experimentation, reflection, feedback, and institutional support before it becomes classroom practice. The coexistence of high flexibility ratings with high workload ratings exposes what they call a common misconception: that accessibility translates into ease of participation. For working teachers, flexible delivery shifts rather than removes the burden of allocating time, sustaining motivation, and balancing coursework against teaching and administrative duties. Barriers, they conclude, are organizational and temporal as much as technological.

The practical implications point away from demographic targeting and toward the conditions surrounding participation. The authors recommend designing relevant, practice-oriented content, reducing unnecessary workload, fostering collaborative learning communities, and embedding MOOCs within school-based professional development structures that provide recognition, mentoring, and protected time. They are careful to frame these as priorities for program development whose effects require direct evaluation, not proven interventions. The study’s limitations, including its cross-sectional design, self-reported data, and convenience sampling, restrict causal interpretation and generalizability. Future research, the team suggests, should employ longitudinal, mixed-method, and performance-based designs to examine whether MOOC participation actually changes instructional practice, how perceptions evolve over time, and which institutional conditions sustain engagement. For now, the message for education systems betting on MOOCs at scale is clear: teachers are ready and willing, but the surrounding ecosystem must catch up.

Subject of Research: Indonesian teachers' perceptions of massive open online courses for continuing professional development

Article Title: Indonesian teachers’ perceptions of MOOCs for digital professional development

Article References: Iqdami, M. N., Haryono, Astuti, T., & Widhanarto, G. P. (2026). Indonesian teachers’ perceptions of MOOCs for digital professional development. Discover Education, 5(1), Article 1080. https://doi.org/10.1007/s44217-026-02170-w

Image Credits: AI Generated

DOI: 10.1007/s44217-026-02170-w

Keywords: MOOCs, teacher professional development, Indonesia, online learning, continuing professional development, survey research, institutional support, workload, teacher perceptions, digital learning, Discover Education, statistical analysis

The core problem the researchers set out to solve is one that has dogged educational recommender systems for years. Most existing methods lean heavily on student interaction data, such as which videos a learner watched or which exercises they completed, and then suggest similar or popular items. That strategy works reasonably well for entertainment streaming, where the cost of a bad suggestion is low. In education, however, the order and structure in which material is encountered matters enormously. Watching an advanced lecture before mastering its prerequisites can leave a student confused, while a well-sequenced pathway can accelerate understanding. Interaction data alone, the authors argue, ignores the intricate dependency networks among learning resources that directly shape the effectiveness of knowledge acquisition, and it fails to model an individual learner’s abilities and goals.

To bridge this gap, the team proposed what they call a unified learning resource recommendation method, or ULRRM. The central innovation is the use of conceptual graphs as an intermediary framework that unifies resource representations across different levels of granularity. In practical terms, this means that a single lecture video, an exercise set, and an abstract knowledge concept can all be expressed within one coherent mathematical structure. Rather than forcing the recommendation engine to choose between recommending fine-grained items or coarse-grained concepts, the graph acts as a common language in which both can coexist, allowing the system to reason fluidly across scales.

The first pillar of the method is a resource dependency graph. This structure encodes the topological constraints of the resource space by capturing conceptual dependency relationships, essentially mapping which pieces of content build on which. When a learner engages with a resource, the graph guides the system toward resources that depend on or extend the concepts just covered, enabling what the researchers describe as resource-dependent learning. The effect is analogous to a well-designed curriculum: the system knows that a student who has just grasped the basics of derivatives is better served by materials on differentiation rules than by a jump into multivariable calculus. By encoding these prerequisite-style relationships directly into the recommendation process, ULRRM ensures that suggestions respect the logical architecture of the subject matter.

The second pillar addresses the temporal dimension of learning. The researchers constructed a local-global dual view built from session history, allowing the model to capture both short-term behavioral patterns and the evolution of long-term interests. The local view focuses on what a learner is doing right now, within a single study session, which is often the strongest signal of immediate need. The global view aggregates behavior across longer horizons, tracing how interests and abilities develop over weeks or months. Combining the two enables the system to recommend not just a single next item, but coherent sequences of learning resources, a capability the authors describe as recommending learning resource sequences that incorporate multidimensional graph information.

Technically, this dual-view design draws on a lineage of session-based recommendation research, which has evolved from recurrent neural network approaches through graph neural networks and self-attention architectures. Earlier educational recommenders often borrowed these techniques wholesale from e-commerce and media streaming, where the goal is simply to predict the next click. The Beihang-led team adapted the machinery to the educational context by weaving in the dependency structure, so that the sequence model is never free to suggest an item whose conceptual prerequisites have not been met. The graph information thus acts as both a guide and a constraint, shaping the embedding space in which learner behavior is interpreted.

The value of this approach becomes clearer when contrasted with the dominant paradigms in the field. Collaborative filtering, the workhorse of classical recommender systems, infers preferences from the behavior of similar users, but it struggles with cold-start learners and says nothing about whether two resources are logically related. Knowledge-tracing models, which estimate a student’s mastery of individual concepts, capture ability but often treat resources as interchangeable instantiations of those concepts. Graph-based course recommenders have begun to exploit prerequisite relations, yet they typically operate at a single granularity. ULRRM’s contribution is architectural: by unifying items, concepts, and dependencies in one graph-based representation, it integrates insights that previously required separate systems.

To validate the method, the researchers conducted extensive experiments on real datasets drawn from online learning scenarios. They benchmarked ULRRM against a range of widely recognized baseline approaches, including session-based neural models, knowledge-graph-enhanced recommenders, and graph convolution methods designed for educational data. The evaluation used standard metrics commonly applied in sequential recommendation research, which measure how well a system places the genuinely useful next item near the top of its ranked list. Across these metrics, the proposed method consistently surpassed the baselines, providing empirical evidence that the multidimensional graph integration translates into measurably better recommendations rather than merely a more elegant theoretical framework.

The implications extend beyond academic benchmarks. For the operators of massive open online courses and other digital education platforms, better sequencing could translate into lower dropout rates, a persistent challenge documented across the MOOC literature. For individual learners, a recommender that understands prerequisite structure behaves less like a content feed and more like a patient tutor, steering students along pathways that build competence step by step. The work also arrives at a moment when large language models are being explored for educational personalization, and the authors’ graph-centric framework offers a complementary strategy: rather than relying solely on the linguistic knowledge of foundation models, it grounds recommendations in the verifiable structure of the curriculum itself.

The research, published as an open-access article in Frontiers of Digital Education with support from the National Natural Science Foundation of China and several Chinese education research programs, arrives at a moment when the volume of online educational content continues to grow explosively. As the authors note, personalized learning resource recommendation exists precisely to alleviate the information overload this growth creates. By encoding the dependency networks among resources, modeling both the moment-to-moment and month-to-month texture of learner behavior, and unifying representations across granularities through conceptual graphs, ULRRM offers a template for the next generation of educational AI. If such systems mature from the lab into production platforms, the frustrating experience of wandering a digital library without a map may give way to something closer to having a knowledgeable guide at one’s side, one that knows not only what to show next, but why it belongs there.

Beyond the headline results, the study sits within a broader research conversation about how prerequisite relationships can be extracted and exploited. Prior work has explored measuring prerequisite relations among concepts in MOOCs using natural language processing techniques, and frameworks have been proposed to capture dependencies between introductory and advanced courses in higher education. By building a resource dependency graph on top of such conceptual structure, ULRRM connects these strands of research with modern sequential recommendation, suggesting a path by which curriculum knowledge curated by educators can be made computationally actionable.

The emphasis on modeling individual learning ability also echoes developments in knowledge tracing, where models estimate a student’s evolving mastery from their answer histories. Context-aware attentive knowledge tracing and graph-based knowledge tracing have shown that representing relationships among concepts improves estimates of proficiency. ULRRM’s dual-view design complements this line of work: rather than diagnosing mastery in isolation, it folds ability signals into the recommendation process itself, so that the sequences offered to a learner reflect both what they have engaged with recently and how their interests have developed over time.

The experimental design reflects standard practice in sequential recommendation research, where evaluation typically measures ranking quality, rewarding systems that surface genuinely useful items near the top of a list. The baselines compared against ULRRM span the field’s recent history, from session-based neural models to knowledge-graph-enhanced and graph convolution approaches, which strengthens the claim that the gains stem from the multidimensional graph integration rather than any single architectural choice.

Several open questions remain for future work. The method’s reliance on conceptual dependency relationships presumes that such structure can be identified accurately, and the quality of the dependency graph will likely bound the quality of recommendations. Scaling the approach to platforms with millions of heterogeneous resources, and validating its effects on actual learning outcomes such as completion and mastery rather than ranking metrics alone, represent natural next steps. The authors note that all data analyzed in the study are included in the published article, which may help other groups reproduce and extend the results as graph-based educational recommendation continues to mature.

Subject of Research: A unified learning resource recommendation method that integrates multidimensional graph information, including resource dependency graphs and dual-view session modeling, to personalize online learning.

Article Title: A Unified Learning Resource Recommendation Method Integrating Multidimensional Graph Information

Article References: Li, S., Liu, X., Zhou, X., Tang, X., & Pu, J. (2026). A Unified Learning Resource Recommendation Method Integrating Multidimensional Graph Information. Frontiers of Digital Education, 3(2), Article 18. https://doi.org/10.1007/s44366-026-0092-0

Image Credits: AI Generated

DOI: 10.1007/s44366-026-0092-0

Keywords: learning resource recommendation, online education, graph neural networks, conceptual graphs, resource dependency, sequential recommendation, session-based recommendation, representation learning, personalized learning, MOOCs, knowledge tracing, educational data mining