The article provides a comprehensive theoretical substantiation and analysis of the practical aspects of implementing artificial intelligence (AI) and educational simulators within the higher education system. The relevance of the study is driven by the urgent need to bridge the gap between traditional academic knowledge and the rapidly evolving requirements of the global digital labor market. The author explores the didactic potential of AI-driven simulators as sophisticated tools for modeling real-world professional scenarios, offering a risk-free environment for the systematic development of students' hard skills and decision-making abilities. The research identifies the critical role of mobile applications in constructing flexible educational trajectories and ensuring seamless, ubiquitous access to interactive learning content. A comparative analysis between traditional teaching models and innovative simulation-based approaches is presented, highlighting the advantages of automated feedback and adaptive learning paths. The study details the architectural components of an intelligent educational ecosystem, including cognitive simulators, analytical modules based on Big Data, and mobile interaction platforms. Significant emphasis is placed on the transformation of the educator's role, shifting from a primary source of information to an architect and moderator of a high-tech learning space. This transition necessitates the continuous development of teachers' digital competencies to manage AI-integrated processes effectively. The findings prove that the synergy of artificial intelligence, mobile technologies, and simulation modeling fosters deep personalization of the learning process. This integration not only enhances student engagement through gamification and SMART-structured tasks but also significantly improves the overall quality and efficiency of professional training for future specialists in various fields, from information technology to physical education and sports. The study concludes that the future of higher education lies in the harmonious balance between human pedagogical expertise and advanced algorithmic support.

Purpose This study aims to examine whether a structurally adaptive e-learning path can improve the efficiency of large-scale digital skills training while maintaining comparable learning outcomes relative to a regular, non-adaptive path. Specifically, it investigates (1) differences in time on task and learning outcomes between adaptive and regular conditions across seven European Computer Driving Licence (ECDL)/International Certification of Digital Literacy (ICDL) modules and (2) whether the two instructional conditions differ in their underlying performance structure and student performance profiles. Design/methodology/approach The study involved 2,064 upper-secondary students enrolled in seven online ECDL/ICDL courses delivered via Moodle. Students were assigned to an adaptive path or a regular full-content path through controlled randomisation stratified by school year and gender. In the adaptive condition, pre-test results unlocked only the learning activities needed to address identified knowledge gaps; in the regular condition, all activities were mandatory. Learning efficiency was assessed through time saved, time attended and topics skipped, while learning effectiveness was assessed through first-attempt pass rates and post-test scores. For structural analyses, we also computed a composite performance index combining achievement and time efficiency, and used confirmatory factor analysis, multi-group modelling and TwoStep clustering to examine latent performance patterns. Findings The adaptive path produced substantial efficiency gains across all seven modules, markedly reducing instructional time and content exposure. Learning effectiveness results were more mixed: outcomes were broadly comparable across several modules, the adaptive group performed significantly better in Computer Essentials and the regular group performed significantly better in Word Processing and Spreadsheets. Overall, the findings indicate a strong efficiency advantage for the adaptive condition without uniform improvement in effectiveness. The course-level performance indicators loaded on a common latent performance factor, although the relative contribution of individual courses differed across conditions. Cluster analysis identified three performance profiles, including a smaller sub-group of consistently low-performing learners present in both conditions. Research limitations/implications The study is limited to one national context, one learning management system and one digital skills curriculum. Pre- and post-assessments reused the same item pools, which may have introduced practice effects. In addition, the performance index and clustering analyses were restricted to successful completers, so these structural findings should be interpreted as applying to completers rather than to the full randomised sample. Future research should test similar designs in other subject domains, including behavioural and motivational measures, and explore richer adaptive logics for practice-oriented modules. Practical implications The findings suggest that structurally adaptive learning paths can meaningfully reduce time-to-mastery in large-scale digital training. However, the mixed effectiveness results across modules indicate that adaptive sequencing should be calibrated to course characteristics and complemented, where needed, by additional pedagogical support for lower-performing learners. Social implications By increasing the efficiency of school-based digital skills training, adaptive e-learning may help education systems deliver certification-oriented programmes under realistic time and resource constraints. Originality/value The study contributes large-scale evidence on a scalable form of structural adaptivity implemented in an authentic school context. Its main contribution lies not in proposing a novel adaptive algorithm, but in testing whether pre-test-based adaptive sequencing can reduce time-to-completion at scale while preserving broadly comparable outcomes, and in linking this question to latent-variable modelling and learner-profile analysis.

This study explores the use of artificial intelligence technology to build an adaptive learning system to promote the innovation of college English teaching. By integrating natural language processing, deep learning, and recommendation algorithms, a multidimensional adaptive learning framework for college English is constructed that provides personalized support in various skills. By designing comparative teaching experiments and collecting learning behavior data, the effectiveness of the model is systematically evaluated. Experiments show that the system can significantly improve the utilization rate of teaching resources, learning efficiency, and interactive experience. Students who use the system have made remarkable progress in vocabulary mastery, writing ability, and oral expression, and their satisfaction is high. Students with different learning styles and levels can get adaptive learning paths and real-time feedback. The adaptive learning system driven by artificial intelligence shows good applicability and popularization potential in college English teaching.

Artificial Intelligence (AI) has emerged as a transformative force in language education offering innovative solutions for personalized instruction, real-time feedback, and inclusive learning environments. While several systematic reviews have explored AI in language learning broadly, few have specifically targeted generative AI using a combined Systematic Literature Review (SLR) and Bibliographic Coupling Analysis (BCA). This study fills this gap by synthesizing findings from 19 peer-reviewed articles published between 2020 and 2024 in the Web of Science database, using a dual approach that combines a SLR and Bibliographic Coupling Analysis to explore the educational potential and thematic development of AI in language learning. The review examines how AI facilitates adaptive learning paths, enhances language skills development, supports assessment and teacher roles, improves accessibility for diverse learners, and raises critical ethical and cross-cultural considerations. Using the PRISMA framework to guide the selection and synthesis process, and bibliographic coupling to identify intellectual linkages, the analysis reveals six main research clusters: personalized and adaptive learning, language skill enhancement, AI-driven assessment, inclusivity and accessibility, ethical and critical engagement, and system usability. According to the literature, AI supports learner autonomy, promotes engagement, and addresses various learner needs, although challenges such as digital inequality, algorithmic bias, and over-reliance on technology persist. In alignment with Sustainable Development Goal (SDG) 4: Quality Education, this study underscores the importance of inclusive, ethical, and learner-centred AI integration. Future research should address the long-term impacts of AI in education, ensure equitable access, and balance technological advancement with pedagogical integrity. This review provides practical recommendations for integrating generative AI into language classrooms, highlights the pedagogical opportunities and challenges associated with AI adoption and outlines future research directions related to long-term learning outcomes and equitable AI implementation.

ABSTRACT With the pervasive adoption of intelligent learning assistants, the demand for genuinely personalised adaptive learning paths is accelerating. However, current path-construction approaches are largely shaped by teacher perspectives and static learner profiles, often overlooking the genuine needs of learners. To address this gap, this study adopts a user-centred approach. Using grounded theory, we systematically analysed Zhihu discussions related to ‘AI and personalised learning’, supplemented by in-depth interviews. Our findings reveal four key modules that drive the dynamic evolution of learning paths: initial path matching, dynamic path adaptation, human-AI collaborative support, and outcome-driven optimisation. These modules interact to form an integrated system for constructing and sustaining personalised adaptive learning paths. This study proposes a novel mechanism model linking technological design with authentic learning experiences and offers practical insights for educators and institutions to embed AI into instruction, ultimately enhancing learning personalisation and adaptation.

This study presents a case from Chongqing, China, on the design and implementation of an AI-supported professional development system for preschool teachers. Grounded in adult learning theory and the Technological Pedagogical and Content Knowledge framework, the system integrates knowledge graphs and deep reinforcement learning to generate adaptive learning paths. Using a quasi-experimental design with 120 teachers, the study compared experimental and control groups in remote training contexts. Results show significant improvement in instructional design, reflection, and collaboration, along with reduced intra-group disparities. Teachers reported high satisfaction with the interpretable recommendation panel, indicating that AI can bridge theory and practice while promoting sustainable digital transformation in preschool teacher development.

This study examined the effects of integrating artificial intelligence (AI) tools into informal digital learning of English (IDLE) to enhance cognitive and non-cognitive skills, as well as listening and speaking proficiency among English as a Foreign Language students. A sample of 120 Egyptian university students participated in a mixed-methods design that consisted of a questionnaire, pretests and post-tests for listening and speaking skills and semi-structured interviews. Quantitative data were analysed using descriptive statistics, t tests and mixed analysis of variance, while qualitative responses were thematically explored. The findings revealed significant advancements in cognitive skills, including the regulation of attitudinal needs, goal commitment, resource allocation and metacognitive skills, as well as enhanced non-cognitive skills. However, social connections via AI were found to be less impactful, with many students reporting limited authentic interactions. While AI-driven IDLE significantly enhanced speaking proficiency, listening skills showed more modest gains, suggesting differential effects of AI on productive versus receptive skills. Despite technical challenges, AI-based IDLE demonstrated potential for personalising learning. Future research should address these challenges while focusing on bridging the gap between informal digital learning and real-world language use. Implications for practice or policy: Educators should integrate AI tools into blended learning models, combining AI-driven practice with real-world communicative opportunities to bridge the gap between simulations and authentic language use. Developers must prioritise customisation in AI tools, such as adaptive learning paths and realistic conversation practice, to address diverse learner needs effectively. Policymakers and administrators should invest in resolving technical barriers (e.g., speech recognition accuracy, Internet reliability) to optimise AI tool effectiveness and user experience.

Introduction Peer-learning recommendation remains an open challenge in e-learning systems, as most existing approaches—such as matrix factorization and neural collaborative filtering—rely on static interaction patterns. These methods often ignore contextual information including learner roles, content difficulty, and temporal engagement behavior. As a result, they struggle to form meaningful peer groups or provide adaptive learning paths that align with pedagogical needs. Methods To address these limitations, we propose a hybrid context-aware peer learning recommender that integrates collaborative filtering with interaction-based clustering. The framework incorporates adaptive peer group formation using multiple loss functions and multifactor BERT embeddings to capture content semantics. In addition, learner-specific characteristics such as difficulty level, job role, and software skills are explicitly modeled. These contextual and semantic features are dynamically used to cluster learners and generate personalized peer recommendations. Results and discussion Experiments conducted on an e-learning dataset demonstrate that the proposed model significantly outperforms sequential baseline approaches, as well as traditional matrix factorization and neural collaborative filtering models. The hybrid approach achieves an accuracy of 0.80, precision of 0.80, recall of 0.06, and an F1-score of 0.11. These results indicate improved personalization and contextual relevance in peer recommendations, enabling more adaptive and pedagogically suitable peer learning experiences.

To address the severe surface imperfections induced during ultrafast pulsed laser fabrication of fused silica microfluidic chips, a high-precision CO2 laser polishing strategy based on shallow-layer melting and reflow was employed. This method enables localized melting within an extremely thin surface layer, effectively smoothing the topography without altering the original microstructure geometry. An L9(33) orthogonal experimental design was conducted to systematically investigate the influence of key parameters on polishing quality, identifying defocus distance as the dominant factor affecting surface roughness, followed by scanning speed and laser power. The optimal parameter combination was determined to be a laser power of 8 W, a defocus distance of 6 mm, and a scanning speed of 5 mm/s. Furthermore, an overlap rate between 38% and 63% was found to ensure sufficient fusion without excessive remelting, with the minimum surface roughness of 0.157 µm achieved at a 50% overlap rate. Based on the optimized parameters, adaptive scanning paths were designed for different functional units of a fused silica microfluidic chip. Surface characterization demonstrated that the surface roughness was remarkably reduced from 303 nm to 0.33 nm, meeting optical-grade surface quality requirements.

BackgroundBiomedical research studies are increasingly using digital tools to enroll, recruit, and collect data from participants. However, variability in digital literacy and technological acceptance can be challenging for recruitment from groups traditionally underrepresented in research, including those served by Federally Qualified Health Centers.ObjectiveThis study aimed to (1) measure participant accessibility and comfort with digital platforms and (2) examine the interrelation of technology access, digital literacy, and support preferences during enrollment and data submission.MethodsA cross-sectional analysis was conducted using enrollment data from Federally Qualified Health Centers participating in the All of Us Research Program. Participants had the option of High-Touch (staff-assisted) or Low-Touch (self-directed) support for enrollment and survey completion. Survey items assessed internet access and technology comfort, while support type was recorded by the research staff based on participants’ actual selection. Logistic regression models evaluated relationships between technology access, comfort, and enacted support type, while controlling for age, consent language, and education, as well as race and ethnicity.ResultsThe analytic sample included 605 participants. The majority reported access to the internet (539/605, 89.1%) and felt comfortable with technology (448/605, 74.1%). In the group requesting High-Touch support (n=346), 14.5% (n=50) reported no internet access, and 31.5% (n=109) felt uncomfortable with technology. In the group requesting Low-Touch support (n=259), 6.2% (n=16) had no access to the internet, and 3.9% (n=10) reported feeling uncomfortable (P<.001). In the adjusted models, much greater comfort with technology was significantly correlated with reduced odds of requesting High-Touch support (comfortable: adjusted odds ratio 0.118, 95% CI 0.055‐0.255 and neutral: adjusted odds ratio 0.212, 95% CI 0.077‐0.587), but internet access was not significantly correlated.ConclusionsThe strongest predictor for support preference for digital enrollment among the participants was their comfort with technology rather than access alone. These findings illustrate the significance of participant-centric design methods coupling adaptive support paths, mixed enrollment strategies, and individualized onboarding methods aligned with digital confidence to promote equitable engagement in precision health research.

Las herramientas SIG de lógica difusa permiten valorar el paisaje en su conjunto, más allá de la rígida conectividad propia del análisis de redes, para seleccionar rutas discriminando flexiblemente temáticas y niveles de esfuerzo o confort climático estacional deseados. Este artículo propone un método de selección de rutas acordes a las preferencias temáticas de los usuarios utilizando indicadores cualitativos del carácter predominante de las rutas, naturalistas o culturales; y parámetros cuantitativos para seleccionar las pendientes y temperaturas medias estacionales deseadas. Esta información reduce la incertidumbre de los usuarios y, con ella, la vulnerabilidad de territorios con componente turística. El análisis de los datos generados por los usuarios abre la posibilidad de implementar un SIG participativo susceptible de sustentar la construcción social del paisaje cultural.

This study proposes a novel approach for physical education (PE) that integrates kinect motion tracking, deep learning, and context personalisation. The system combines real-time feedback and adaptive learning paths to optimise student participation, motivation, and physical skill development. An ablation study was conducted to compare the effectiveness of the full system with three other configurations: kinect-only motion tracking, kinect with context personalisation, and kinect with deep learning. The experimental results indicate that the full system, which combines all three components, significantly outperforms the other configurations in terms of motivation, physical performance improvement, and engagement. Specifically, the full system achieved the highest improvement in skill development (90%), engagement (98%), and motivation, suggesting that the combination of kinect motion tracking, context personalisation, and deep learning is most effective for enhancing PE outcomes. This research contributes to the digital transformation of physical education. It provides a new pathway to leverage technology for improving both student motivation and performance.

This research examines how artificial intelligence (AI) can be used to improve ethical leadership and lifelong learning programs in organizations. Data was collected using a convergent mixed-method design by carrying out 20 semi-structured interviews, three focus groups, and 200 survey participants in corporate, educational and governmental institutions that utilize AI in leadership development. Thematic analysis was employed in the analysis of qualitative data and the survey data were discussed statistically to determine the impact of AI on ethical decision-making, transparency, and lifelong learning. The evidence indicates that AI can enhance the ability of leaders to make transparent, accountable, and responsible decisions and provide personalized and adaptive learning paths to support ongoing professional development. Algorithms, less human discretion, and the lack of trust in AI results were recognized as the ethical issues by the participants. The study also builds on the theory by harmonizing AI, ethical leadership, and lifelong learning into one framework and provides a practical recommendation to organizations aiming to adopt AI in a responsible manner. These lessons provide the promise of AI in advancing inclusive, ethical, and adaptive leadership and the necessity to govern, be transparent, and be AI literate.

The Artificial Intelligence (AI) applied to dance pedagogy is a revolutionary innovation in the process of movement competencies instruction, rehearsal, and assessment. This paper assesses how the AI-based instructional systems affect dance education by comparing them to the existing learning theory, principles in motor learning, and the kinesthetic intelligence. The proposed framework, based on constructivism, embodied cognition, and experiential learning, will make AI an intelligent educational companion as opposed to a substitute of human instructors. To examine the movement accuracy, time coordination, and quality of expression, AI technologies that consist of computer vision-based pose estimation, machine learning-based skill evaluation, and sensor-based motion capture are used. A mixed-methodology is used which implies dance students and teachers of different styles and levels of expertise. The data is gathered by video recordings, wearable devices, questionnaires, and semi-structured interviews, which allows quantifying and qualifying it. The AI models are trained to give personalized feedback, identify the movement errors, and give adaptive learning paths, both in real-time and offline learning. Comparative analysis demonstrates that AI-assisted learning results in definitive conclusions of movement accuracy, time, and expressiveness in comparison to conventional pedagogy. Moreover, learners are found to be more engaged, they are more self-aware, and autonomous in practice sessions. The results show that AI systems that are pedagogically aligned can be used to facilitate reflective learning, promote individualized instruction, and supplement teacher-led dance learning.

Students today find it difficult to have access to resources that closely match their skills and career expectations with the educational scenario changing almost every day. E- Vidya attempts to fill this gap by bringing together advanced AI-enabled recommendations, learning through experiences and effective career guidance. And it is more than just delivering content to the students; the platform has a structured module, previous years sample papers, code exercise and also very detailed placement guide all as per student’s requirement based on what each student intend to achieve. Starting from teaching and enhancing their tech skills, learning new certifications or preparing for placements. Motivation tools like bonuses, badges, and leaderboards drive students to want to continue and push themselves in a nice space. The forum for teachers, interactive courses and quizzes all ensure that a lively and engaged interaction takes place, in which a real community is actively participating. E-Vidya ultimately serves as a reliable mentor in the journey of a learner connecting the traditional gap between classroom instruction and modern electronic teaching, so that kids feel readier, more self-assured, and more inspired for their future.

This study explores how deep learning (DL) technologies, particularly those integrated into tools such as automatic speech recognition (ASR), text-to-speech (TTS), and AI-powered pronunciation applications, can transform speaking instruction by introducing a new pedagogical paradigm. Through a qualitative, literature-based methodology, this research analyzes thirty peer-reviewed studies published between 2020 and 2025. The findings reveal that DL tools significantly enhance learners’ pronunciation, rhythm, and fluency through real-time feedback and adaptive learning paths. In addition, these tools contribute to affective gains by reducing speaking anxiety and increasing learner motivation. However, challenges such as algorithmic bias, limited effectiveness in pragmatic discourse, and a lack of teacher training remain significant barriers to full integration. The discussion highlights the shifting role of TEFL instructors from content deliverers to facilitators of technology-enhanced learning environments. The study also emphasizes the importance of considering ethical implications and ensuring data privacy when implementing DL applications. Recommendations are provided for educators, institutions, and researchers to foster effective, equitable, and sustainable use of DL in language education. By bridging the gap between AI technologies and pedagogical practice, this paper proposes a forward-looking framework that positions deep learning not merely as a tool but as a transformative force in TEFL speaking instruction.

Abstract This study aims to explore how startups can enhance resilience through business model transformation (BMT) driven by data-driven methodologies during crises. Organizational resilience, defined as the ability to adapt, recover, and thrive amidst adverse conditions, has become a strategic imperative in an era marked by repeated global disruptions. While previous research has focused on adaptive and absorptive paths to resilience, our work highlights the synergistic potential of integrating these approaches with data-driven growth, in the specific and under-researched context of platform-based startups. Adopting a qualitative research design, we conduct a multiple case study of four platform-based startups from diverse sectors. Our findings reveal that data-driven growth facilitates continuous experimentation, real-time learning, and agile decision-making, enabling organizations to identify new market opportunities, address customer pain points, and pivot their value propositions effectively. By fostering dynamic capabilities, such as sensing, seizing, and reconfiguring resources, this approach enhances both short-term adaptability and long-term competitiveness. The study contributes to the resilience, business model, and growth literature by advancing our understanding of how data-driven methodologies can act as a catalyst for BMT and organizational resilience, offering actionable insights for both theory and practice. Our framework underscores the strategic importance of integrating growth hacking principles into business model innovation to build robust organizational resilience in an increasingly turbulent environment.

Occlusion-aware dual-memory transformer for unsupervised UAV video stabilization with adaptive inference paths

This paper investigates a communication-constrained unmanned aerial vehicle (UAV) pickup and delivery system for continuous multi-period operations. To ensure real-time control updates between UAVs and the ground server, a minimum communication rate requirement is imposed throughout each mission. The objective is to minimize the average mission completion time of multiple rotary-wing UAVs while satisfying mobility, payload, safety, and communication constraints. The resulting mixed-integer nonlinear programming problem, involving binary pickup/drop-off decisions, trajectories, and variable time-slot durations, is mathematically intractable. To address this, a successive convex approximation framework combined with a penalty convex–concave procedure is developed, enabling iterative convex reformulation and convergence to a near-optimal binary-feasible solution. Simulation results demonstrate that the proposed algorithm efficiently generates collision-free trajectories and adaptive flight paths that maintain reliable communication links, outperforming baseline strategies in terms of completion time and coordination efficiency under communication constraints.

Abstract Abstract— The rapid growth of digital learning platforms has created a strong demand for intelligent systems capable of delivering personalized learning experiences. Traditional e-learning environments often rely on generic content delivery, which fails to adapt to the diverse learning styles, performance levels, and behavioral patterns of students. To address these limitations, this paper presents a comprehensive review of intelligent tutoring systems and proposes a hybrid machine learning–based personalized content recommendation framework. The proposed model integrates collaborative filtering, content-based filtering, and learning-style classification with an ensemble ranking mechanism to generate adaptive learning paths for individual students. The hybrid approach overcomes cold-start issues, enhances recommendation accuracy, and supports continuous learner profiling through real-time feedback analysis. This review highlights the strengths and limitations of existing approaches, identifies key research gaps, and demonstrates how hybrid ML models can significantly improve the effectiveness of AI-driven tutoring systems. The findings suggest promising applications for K–12, higher education, and skill-development platforms, paving the way for next-generation personalized AI tutors. Key Words: Intelligent Tutoring System, Personalized Learning, Hybrid Machine Learning Model, Content Recommendation, Collaborative Filtering, Content-Based Filtering, Learner Profiling, Adaptive Learning, Educational Data Mining, AI in Education.