Frontiers in Education: Digital Learning Innovations

BackgroundArtificial intelligence (AI) is increasingly reshaping medical education through personalized learning, adaptive assessments, and advanced simulations. This systematic narrative review synthesizes the theoretical development of AI in medical training, focusing on educational models, frameworks, learning outcomes, and stakeholder considerations.A literature search of PubMed, Scopus, Web of Science, and Google Scholar (January 2000–March 2025) identified 1,288 records, of which 48 studies met the inclusion criteria and were included in qualitative thematic synthesis. No statistical meta-analysis was conducted due to methodological heterogeneity.ResultsFive major AI domains emerged: Intelligent Tutoring Systems, Simulation-Based Medical Education, Adaptive Learning, Generative AI, and Explainable AI. These domains align with established instructional theories and contribute to improved engagement and learning efficiency. However, concerns persist regarding learner deskilling, academic integrity, and algorithmic bias. AI integration influences multiple stakeholders, including trainees, educators, clinicians, policymakers, and patients. The field has progressed from rule-based approaches to data-driven machine learning models, enabling personalized instruction. Responsible implementation necessitates addressing pedagogical, ethical, and practical challenges, while also reducing the global digital divide.ConclusionsThis systematic review provides guidance for educators, researchers, and policymakers on integrating AI effectively and ethically into medical education.

IntroductionExpertise in video games is commonly stratified using self-reported experience, which is prone to bias and measurement error. This study introduces a behavioral observation approach to provide a more objective and psychometrically sound method for differentiating expertise. Specifically, the Behavioral Observation Matrix-Proxemics (BOM-Proxemics) was developed to assess expertise through observable in-game behaviors grounded in digital proxemics theory.MethodsUsing recorded gameplay from Apex Legends, 102 players were evaluated as novice, competent, or expert based on in-game observed behaviors. The BOM-Proxemics measured behaviors across four domains: Spatial Positioning, Spatial Realization, Spatial Appropriation, and Spatial Interactivity. Reliability and validity were evaluated through inter-observer agreement, internal consistency, and criterion-related analyses. Predictive relationships were examined using ordinal logistic regression, and group differences were analyzed using ANOVA.ResultsThe BOM-Proxemics demonstrated strong reliability and validity, including excellent inter-observer agreement and high internal consistency. Total scores were a significant positive predictor of in-game rank. Among subscales, Spatial Positioning, Spatial Appropriation, and Spatial Interactivity significantly predicted rank, while Spatial Realization did not. Significant differences in scores were observed across all expertise groups, with pairwise comparisons indicating clear separation between novice, competent, and expert players.Discussion/ConclusionFindings support the BOM-Proxemics as a psychometrically sound, behavior-based measure of video game expertise. The results demonstrate the instrument's ability to differentiate expertise levels and predict performance outcomes, offering a viable alternative to self-report measures. Implications include advancing expertise research through direct observation and extending behavior-based assessment approaches to other complex digital environments.

Foreign language speaking anxiety remains a major affective barrier in EFL oral communication, particularly in classroom speaking activities. This study investigated the effects of an AI-supported speaking intervention on EFL learners' oral proficiency, speaking anxiety and learner's perception across different task types. Adopting a mixed-method quasi-experimental design, 83 Chinese university students participated in an AI-supported speaking intervention, with follow-up interviews conducted with 12 participants. Quantitative results showed that the experimental group outperformed the control group in oral proficiency (t = 2.81, p = 0.006). No significant difference was found in overall foreign language speaking anxiety between groups. Task-based analyses revealed that AI conversational agent reduced situational speaking anxiety in pair work and presentation tasks (p < 0.001), but not in debate or storytelling. Qualitative findings further showed that learners perceived AI conversational agents as providing scaffolded and adaptive support through personalized feedback, prompting, and repetition in a non-judgmental environment, which facilitated oral development and reduced performance-related anxiety. Overall, the findings suggest that AI conversational agents support oral performance and selectively alleviate situational anxiety, while their effectiveness remains contingent on task structure and pedagogical design.

Notetaking with digital devices during asynchronous online learning remains controversial. This study investigated how three writing devices—ink pens, digital pens, and laptops—interact with active (verbatim) vs. constructive (question) notetaking strategies during video lectures. During a within-design laboratory experiment, EEG data was recorded while 33 undergraduate students took notes for learning sessions using different notetaking devices and strategies, followed by immediate post-tests. Time-frequency analysis revealed significant differences in theta, alpha, beta, and gamma band power across devices, and significant interaction effects in beta, gamma, and theta/beta values. Both pen types showed higher alpha, beta, and gamma power and lower theta/beta ratios compared to keyboards, particularly in occipital regions associated with sustained visual attention. However, interaction effects indicate the importance of notetaking strategy, and immediate post-test performance showed no significant differences across conditions. The findings suggest that notetaking media influence learning processes and attention sustainment differently, though immediate performance outcomes remain similar. This has implications for designing asynchronous online learning environments and guiding notetaking practices in those settings.

The growing presence of artificial intelligence (AI) across educational and workplace environments is reshaping how learners encounter tasks, interpret feedback, and navigate uncertainty. To understand these changes, this manuscript grounds AI's influence in theories of self-regulated learning (SRL), which conceptualize learning as a cyclical process of planning, monitoring, strategic adjustment, and reflection. Rather than replacing these processes, AI reshapes the conditions under which they occur by making some cues more visible, introducing new forms of guidance, and occasionally preempting difficulty before learners have an opportunity to engage with it. These shifts reveal a conceptual gap: although research documents both benefits and risks of AI-mediated support, we lack a framework for understanding how AI participates in learners' regulatory cycles across educational and professional settings without eroding the autonomy that underpins SRL. To address this gap, this article proposes a unified model of AI as a co-regulator within self-regulated learning, grounded in Winne and Hadwin's COPES architecture. The model centers productive metacognitive friction as a mechanism for sustaining learner-driven regulation by structuring how learners encounter challenge and discrepancy. It advances a relationally grounded framework at the level of interactional structure, positioning AI as a co-regulator through five design principles that specify conditions under which AI can support regulatory cycles without displacing learner judgment. These principles are linked to an evaluation architecture that centers autonomy, interpretability, process integrity, and developmental growth as evaluative priorities traced through learner–AI interaction patterns. Implications are examined across educational practice, workplace learning, equity, and governance, and directions for collaborative research and design are outlined to investigate how relationally aligned AI can preserve and strengthen the regulatory processes at the heart of SRL.

BackgroundGamified, AI-enabled digital learning environments (GAI-DLEs), integrating adaptive or generative AI with game-based design, are increasingly used in school science education to support personalized learning. However, consolidated evidence on their effectiveness, implementation models, and regional distribution remains limited, particularly in Central Asia.MethodsFollowing PRISMA 2020 guidelines, a systematic search of Scopus and complementary databases identified peer-reviewed empirical studies published between 2015 and 2025. A total of 1,284 records were identified. After deduplication, 962 unique records were screened at the title and abstract level, and 150 full-text articles were assessed for eligibility. Ultimately, 81 studies met the inclusion criteria and were included in the final synthesis. The included evidence was examined through trend analysis, thematic synthesis, and geographic mapping. The meta-analysis included experimental and quasi-experimental studies from formal school settings. Standardized mean differences were estimated using random-effects models, with subgroup analyses by education level, AI technology type, publication period, study quality, and region.ResultsGAI-DLEs demonstrated a significant positive effect on science learning outcomes compared with non-AI instructional conditions (SMD = 1.01; 95% CI: 0.69–1.33; p < 0.001). Effects were stronger in secondary education (SMD = 1.12; 95% CI: 0.69–1.55) than in primary education (SMD = 0.80; 95% CI: 0.35–1.25). Studies employing adaptive or generative AI systems tended to report larger effect sizes. Evidence regarding student engagement was generally positive but showed substantial contextual heterogeneity.ConclusionGAI-DLEs show consistent potential to improve science learning in school contexts. However, the global evidence base remains geographically imbalanced, with Central Asia substantially underrepresented. Future research should adopt theory-driven and longitudinal designs to examine how specific combinations of AI functionalities, gamification mechanics, and classroom integration strategies produce scalable educational outcomes.

IntroductionVirtual laboratories (vLabs) are increasingly used in engineering education to support preparation for complex experimental work. We report the implementation and exploratory evaluation of a serious-game-inspired vLab that mimics a biomechatronics device, the MyoRobot, within a Master's-level course.MethodsThe Unity-based desktop simulation combines a realistic 3D laboratory, an interactive technical anatomy model, and a pipetting and operation workflow that links virtual actions to plausible force-recording outcomes. In an initial cohort (N = 8), students prepared either using a written manual (n = 4) or primarily the vLab (n = 4). Open-ended questionnaires administered before and after the physical lab assessed perceived preparedness, confidence, engagement, and overall user experience.ResultsvLab users reported improved conceptual orientation and procedural confidence, along with greater independence in handling sensitive equipment. However, they also noted a higher perceived time investment, a reduced novelty effect during the physical lab, and a desire for more specific feedback.DiscussionWe interpret these findings as context-specific insights from a pilot study and derive design implications for realistic vLabs that aim to balance guidance, workload, and authenticity in engineering education.

IntroductionThis study examines heterogeneity in the mediating role of self-efficacy between prior AI training and adoption intentions among pre-service mathematics teachers.MethodsUsing data from 79 pre-service teachers at the University of the Free State, South Africa, Bayesian moderated mediation analysis was employed to assess whether this pathway operates uniformly across demographic subgroups.ResultsFindings revealed pronounced heterogeneity: the indirect effect was strong for female participants (indirect effect = 0.311, P(>0) = 94.8%) but negligible for males (indirect effect = −0.064, P(>0) = 37.8%). Additionally, self-efficacy predicted intentions more strongly among untrained (β = 0.746) than trained teachers (β = 0.195).DiscussionThese results suggest that training may homogenise intention formation and that self-efficacy operates differently across subgroups. The findings challenge uniform models of technology adoption and highlight the need for differentiated, context-sensitive teacher education strategies

AI-based technologies have successfully proliferated across various levels of education, from higher to elementary education. The most apparent implementation of AI in education can be traced directly to higher education, which has a number of potential application areas. Despite its popularity at the higher education level, its application in the context of elementary education remains scarce in the literature. Investigating the challenges associated with implementing AI-based technologies in elementary education is equally important as that of the widely tackled field in higher education. Along this line, this paper intends to explore the challenges brought about by AI in elementary education using the fuzzy best-worst method. A case study in a cluster of elementary education institutions in Cebu City, Philippines, is conducted, and interesting results reveal that stakeholders prioritize addressing the generalizability of the data mining model prior to the actual adoption of AI in elementary education.

IntroductionThe inclusion of a system dynamics course in our medical school curriculum was designed to encourage systems thinking through computational modeling. From anecdotal observations, it soon became evident that something more profound was occurring—rather than simply learning, our students appeared to be constructing knowledge by building computational models in a way that is consistent with Papert's constructionism.ApproachIn the absence of a reliable tool to identify constructionism, we examine the seminal literature of Forrester's system dynamics and Papert's constructionism by extracting key excerpts to look for evidence supporting the hypothesis that computational modeling may constitute a constructionist activity.ObservationsThe literature suggests that there is substantial convergence between the educational approach of constructionism and the activity of constructing models in system dynamics.DiscussionAn examination of the seminal literature suggests that system dynamics modeling has features that are consistent with a constructionist approach. By extension, other approaches such as agent-based modeling also embody constructionist principles, and the expanding integration of artificial intelligence into computational modeling may present opportunities for novel approaches to constructionist learning. Formal real-world educational studies will be required to accumulate empirical learner data in order to confirm the constructionist nature of systems modeling.

This study aims to explore secondary school students’ perceptions of the implementation of an active methodology -Story-Based Learning (SBL)- and the role of technology and artificial intelligence as mediators in fostering its creative phase. The methodology employed follows a descriptive approach, using a survey design and an analysis of the relationships between the observed variables through Pearson's correlation coefficient (r). An experimental situation was designed with a sample of 164 students, of whom 110 took part in a technology-mediated didactic experience applying the method. This group was asked to complete a creative activity to complement the lesson they had received, under the condition that artificial intelligence should not be used in the process. The experimental group, consisting of 54 students, carried out the same activity with the condition of incorporating artificial intelligence into their creative process. The results indicate that the applied methodology is perceived as more engaging and more capable of sparking interest and fostering creativity than traditional methodologies. Regarding the use of Artificial Intelligence, those who abstained from using it felt more creative, both individually and collectively, and more connected to their peers than those who did. However, those who utilized AI perceived a slight improvement in the quality of their final output. These findings suggest that AI should be integrated as an additional voice to stimulate brainstorming and creative debate, rather than as a replacement for them. It is perhaps best suited to the final stages of the process to refine, polish, and enhance the final product.

BackgroundVeterinary anatomy education has traditionally relied on cadaveric specimens and two-dimensional (2D) educational resources, such as textbooks and atlases. However, ethical, logistical, and biosafety constraints increasingly limit access to cadaveric material, particularly for teaching the anatomy of wild and exotic animals. Advances in three-dimensional (3D) modeling and printing, along with the growing availability of digitized anatomical collections, have expanded the range of educational resources available for anatomy teaching. Nevertheless, comparative evidence regarding the educational effectiveness of these alternative approaches remains limited. This study evaluated whether learning animal osteology using 3D-printed anatomical biomodels yields learning outcomes comparable to those obtained with real anatomical specimens and using only 2D educational materials.MethodsA quantitative controlled experimental study was conducted with 261 undergraduate veterinary medicine students, who were categorized according to prior knowledge of osteology into a group with previous anatomy training and a group without prior training. Students were randomly allocated to one of three study methods: 3D-printed anatomical biomodels, real cadaveric specimens, or 2D materials only. All groups received a standardized 45-minute study session, followed by a practical assessment conducted exclusively using real anatomical specimens.ResultsStudents with prior knowledge achieved higher overall scores than those without previous training. Within both subgroups, students who studied using 3D-printed biomodels or real anatomical specimens obtained significantly more correct answers than those who relied exclusively on two-dimensional materials. No significant differences were observed between the 3D biomodel and real specimen groups. Perfect scores were achieved only by students with prior knowledge who studied using 3D biomodels or real specimens. Test completion time and performance did not differ according to gender.Conclusions3D printed anatomical biomodels yielded learning outcomes comparable to those obtained with real anatomical specimens and superior to those achieved using only two-dimensional materials. These findings support the use of three-dimensional biomodels as a viable and effective educational resource in veterinary anatomy education, particularly in contexts constrained by ethical, logistical, or conservation-related limitations, and highlight their potential to expand access to diverse anatomical specimens.

IntroductionThis study investigates how ChatGPT, a generative AI tool, can promote inclusive and sustainable e-learning by enhancing learner empowerment, satisfaction, and engagement.MethodsDrawing on the Technology Acceptance Model (TAM), Self-Determination Theory (SDT), and the Sustainable Education Framework, a structural model was developed to examine the mediating roles of student empowerment and student satisfaction, and the moderating role of inclusivity perception. Data were collected from 350 students across South Korean universities using a cross-sectional survey and analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM).ResultsThe findings indicate that perceived usefulness, perceived ease of use, and accessibility of ChatGPT significantly enhance student empowerment, which in turn increases satisfaction, continuance intention, and sustainable learning outcomes. Furthermore, inclusivity perception strengthens these relationships.DiscussionThe study contributes theoretically and aligns AI-enabled education with SDG-4.

IntroductionAs digital technologies continue to impact our education system, webinars have become an essential way to deliver timely, scalable training, especially in the post-COVID era. Despite widespread use, we still know relatively little about how effective webinars are especially in the field of agricultural education and extension. This study aims to fill that gap by exploring how socio-economic, personal, and learning-related factors shape participants' webinar experiences and outcomes.MethodsData was collected from 415 participants across India through a structured online survey. The impact of webinar was assessed using the first two levels of Kirkpatrick's evaluation model, focusing on participants' reactions and learning. To measure overall effectiveness, Webinar Effectiveness Index (WEI) was developed using the Fuzzy Analytic Hierarchy Process (Fuzzy-AHP) based on five key components like usefulness, lecture quality, knowledge gain, satisfaction, and learning impact. Fuzzy C-Means clustering was applied to identify patterns among learners, supported by correlation analysis to understand how satisfaction and learning outcomes relate. Costs and time requirements of webinars were also taken into account.ResultsOur analysis revealed eight distinct types of learners, each showing different levels of engagement and effectiveness. Some clusters performed consistently well, reporting strong knowledge gains and high satisfaction, while others displayed more varied and less favorable outcomes. We also found clear, positive Correlation between participants' satisfaction, their knowledge gain, and the overall impact of the learning experience. Additionally, webinars proved to be more economical and time-efficient than in-person seminars.Discussion/ConclusionThe study offers a practical, multi-dimensional approach to evaluating webinar effectiveness using soft computing tools. The findings highlight how learner diversity shapes digital learning outcomes and demonstrate the strong connections among satisfaction, learning, and perceived impact. Overall, the study provides useful guidance for designing webinars that are more engaging, inclusive, and cost-effective especially for large-scale capacity-building programs in agriculture and other fields where accessibility and scalability matter most.

IntroductionThe application and influence of artificial intelligence (AI), and specifically Large Language Models (LLMs), in educational processes is widely discussed. However, there remains a gap in research on using LLMs as peer-like contributors in collaborative learning contexts.MethodsThis article reports a mixed-methods quasi-experimental study investigating how positioning ChatGPT as a peer-like feedback provider shapes student-teachers’ learning and collaboration during group lesson-design activities. The study employed a counterbalanced crossover structure for knowledge assessment and a sequential two-task design for authentic artifact production. A total of 102 teachers in training (M_age = 38.87, SD = 8.01), organized into 21 groups, completed two authentic design tasks within a single session.ResultsAcross the session, students progressively adapted to AI interaction, refining how they queried the model and how they evaluated and integrated its suggestions. Results indicate a Post-Withdrawal Sustained Performance (PWSP) effect: improvements observed during AI-available phases were not followed by a detectable decline in the immediately subsequent AI-withdrawn phase within the study timeframe. This pattern was clearest for technology-related knowledge and was consistent with stable artifact quality after AI removal. While ChatGPT support increased efficiency and contributed to technology-focused insights, qualitative evidence also pointed to tensions, including reduced peer-to-peer idea-building in some groups and concerns about creativity.DiscussionOverall, the findings suggest that integrating LLMs as a feedback team-mate can support collaborative design work without immediate post-withdrawal performance costs, particularly when learners are scaffolded to engage critically with AI output rather than accept it unreflectively. These results carry implications for the design of AI-enhanced collaborative activities, highlighting the need to balance AI efficiency gains with sustained opportunities for authentic peer dialogue.

Despite the proven efficacy of corpus linguistics in language education, its adoption in English for Specific Purposes (ESP) classrooms remains limited due to the technical complexity of corpus tools, particularly in non-English departments. Addressing this “usability gap,” this study evaluates the effectiveness of the Digital Corpus-based Instruction (DCI) model, a novel instructional framework that synergizes corpus data, QR-code technology, and Mobile-Assisted Language Learning (MALL). The study aims to determine if providing corpus data through familiar mobile interfaces can improve lexical competence among non-English majors in low-resource settings. Employing a quasi-experimental design with a convergent parallel mixed-methods approach, the study involved 30 non-English-major undergraduates in an Indonesian university setting. Quantitative data from standardized pre- and post-tests, focusing on vocabulary matching, collocation awareness, reading, and professional writing, were triangulated with qualitative insights from classroom observations, Focus Group Discussions (FGDs), and interviews to capture behavioral and cognitive shifts. The results revealed a statistically significant improvement in the experimental group (p < 0.001), with a very large effect size (Cohen's d > 1.93). Significant gains were observed in collocation awareness (from 2.30 to 5.13) and professional writing accuracy (from 7.83 to 10.10). Qualitative findings corroborated these metrics, indicating a pedagogical transformation from passive learning to active, data-driven inquiry and increased professional confidence. This study contributes to the field by demonstrating that DCI effectively reduces the cognitive load of corpus analysis, offering a scalable, pedagogically viable solution for enhancing ESP competence in higher education contexts.

In the era of digital transformation, the application of remote sensing (RS) data in geography education and its effective integration into the teaching and learning process has become one of the pressing educational challenges, particularly in fostering students’ spatial thinking competencies. The purpose of the study is to identify the possibilities of incorporating RS data into school geography, to develop an effective teaching methodology based on its application, and to test it in practice. The research employed a review of international publications, surveys, correlation and comparative statistical analysis, modeling, analysis of regulatory documents, and pedagogical experimentation. The survey involved 124 geography teachers from different regions of Kazakhstan, while the experimental study included 52 eleventh-grade students from one secondary school selected through random sampling (26 students in the experimental group and 26 students in the control group). The survey results revealed that 62% of teachers use satellite imagery during lessons; however, 69% of them utilize RS data primarily for visualization purposes. The results of the pedagogical experiment demonstrated that the average academic achievement of students in the experimental group increased by 18.2%, whereas the improvement in the control group was 3.6%. Furthermore, the use of RS data positively influenced students’ geospatial thinking, research skills, and analytical abilities. The study further established that the structural-content model of RS-based teaching materials developed as a result of the research can contribute to improving existing methodologies. The proposed model is designed to modernize school geography curricula, enhance teachers’ methodological capacity, improve students’ learning outcomes, and promote geospatial literacy. The findings of the study demonstrate that the systematic integration of RS data into school geography education can serve as a scientific and methodological foundation for modernizing educational content, fostering students’ contemporary scientific and technological competencies, and improving the overall quality of geographic education.