PurposeContinual learning models offer a transformative approach to artificial intelligence (AI) in education by enabling systems to incrementally adapt to new tasks and data while preserving previously acquired knowledge. This stands in contrast to static AI systems, which are trained once on fixed datasets and cannot evolve after deployment. In dynamic educational environments where student needs, curricular goals and teaching strategies shift frequently, this adaptability is essential. This article explores how continual learning systems can support human-AI collaboration by functioning as intelligent instructional partners rather than replacements for educators.Design/methodology/approachWe propose two conceptual applications: (1) intelligent tutoring systems that evolve with each cohort while supporting instructor oversight and (2) AI-assisted curriculum co-design tools that analyze longitudinal learning patterns to guide course refinement. These concepts are examined through two case studies situated within engineering education.FindingsWhen embedded in human-in-the-loop frameworks, continual learning systems have the potential to function as adaptive, trustworthy educational technologies. The proposed intelligent tutoring system is designed to support personalized feedback without forgetting prior instructional logic. The proposed curriculum co-design tool is intended to track outcome performance trends across terms and provide evidence-informed insights for curriculum review.Research limitations/implicationsAs conceptual prototypes, the proposed systems prioritize architectural innovation over immediate empirical validation. This intentional abstraction allows for the reimagining of AI's role in education but limits current measurement of efficacy or scalability. The reliance on continual learning assumes future-ready data infrastructures and institutional willingness to co-evolve with adaptive systems, conditions not yet universally met. These limitations are catalytic, not constraining: they invite interdisciplinary research into explainability, governance and human-AI co-design. Realizing the full potential of these models will require iterative development, simulation-rich testing environments and strategic alignment with evolving standards for ethical, transparent AI in education.Practical implicationsEmbedding continual learning into educational AI systems enables sustained instructional adaptability, helping institutions respond to evolving learner needs, curriculum shifts and assessment strategies without retraining models each term. The proposed intelligent tutoring system supports faculty in delivering personalized, context-aware feedback, while the curriculum co-design tool empowers programs to make data-informed revisions grounded in longitudinal trends. These tools can reduce faculty workload, enhance instructional agility and promote evidence-based teaching practices. Institutions that invest in such systems will be better positioned to foster responsive, scalable and learner-centered education, particularly in dynamic disciplines like engineering where content mastery evolves over time.Social implicationsBy developing human–AI partnerships based on continual learning, this framework supports more equitable, transparent, and responsive educational ecosystems. These systems can help reduce achievement gaps by personalizing support based on evolving learner needs, while preserving pedagogical consistency across cohorts.Originality/valueThis work introduces a novel approach to educational AI that emphasizes co-evolution with educators. By embedding continual learning into instructional and curricular tools, this work proposes a foundation for sustainable, transparent and relevant AI-enhanced learning ecosystems. Future directions include implementation within virtual learning environments and broader programmatic assessment.
PurposeThe rapid integration of artificial intelligence (AI) in education presents many opportunities but also raises critical ethical and cultural challenges. This study aims to explore the current application of AI in education in Myanmar and investigates the conflicts between global AI ethical principles and local cultural values.Design/methodology/approachEmploying a mixed-methods research design, the study utilizes quantitative survey analysis alongside qualitative participatory workshops to provide both breadth and depth to the understanding of educators' perspectives.FindingsThe quantitative findings indicate that while most ethical considerations and practical AI implementation patterns remain consistent across different groups, educators' concern for maintaining human autonomy in AI-assisted education grows significantly with their level of professional experience. Furthermore, the qualitative insights highlight a clear gap between internationally developed (often Global North-centered) AI ethics frameworks and the realities of Myanmar's education context. In particular, socioeconomic challenges and local cultural expectations shape how educators understand and respond to AI. Overall, the study argues that AI ethics in education need to be more locally grounded and responsive to national contexts, rather than relying solely on universal frameworks.Originality/valueThis paper fulfils an identified need to study how ethical AI integration can be enabled in under-resourced and underrepresented educational contexts. By centering the voices of Myanmar educators and documenting their unique perspectives, the study advances knowledge on AI ethics in education and provides practical, context-sensitive guidance for educators, policymakers and researchers.
Urban growth, climate pressures, and the global energy transition are driving the urgent modernization of traditional infrastructure into adaptive, intelligent systems. Emerging technologies such as Artificial Intelligence (AI), Augmented Reality (AR), Natural Language Processing (NLP), and the Internet of Things (IoT) provide powerful opportunities, yet their application in energy and building systems is still fragmented. This paper examines how these technologies converge within smart grids and building technologies. It reviews state-of-the-art engineering solutions, evaluates industrial applications, and analyzes persistent challenges including interoperability, trust, and cybersecurity. The study introduces a modular, scalable integration framework that is validated through case studies and highlights innovation pathways such as quantum machine learning and brain–computer interfaces (BCIs). Findings demonstrate improvements in efficiency, resilience, and sustainability across multiple domains, offering a foundation for the development of next-generation infrastructure.
The digital divide is still a major obstacle to fair education and digital collaboration, especially in rural, remote, and underserved areas of developing countries, even though internet infrastructure is growing rapidly around the world. An important flaw in traditional Information and Communication Technologies for Development (ICT4D) models is that internet access is not only scarce in these places, but frequently nonexistent, unstable, or unaffordable. This paper details the design and technical validation of ChipServer™, a novel, patented, self-contained device and platform engineered to facilitate secure, decentralized file sharing and content distribution by allowing users to access, share, and project content without relying on external internet, Wi-Fi, downloadable apps, or even a computer. ChipServer creates its own private WPA-3 secured Wi-Fi network to broadcast content to users in a classroom, lecture hall, or field settings. Most document-sharing programs rely on external internet access or periodic synchronization. This dependence introduces security vulnerabilities and logistical barriers that prevent effective digital teaching and real-time collaboration in low-connectivity or highly secure environments. The system’s robust, browser-based architecture and power efficiency provide a model for highly resilient, autonomous content delivery, demonstrating a viable technical solution to the infrastructural component of the digital divide.
Academic assessments have long relied on written exams as a primary evaluation method, but the increasing prevalence of exam malpractice undermines their effectiveness. Cheating, particularly in time-bound assessments, has become a global concern. Traditional manual invigilation remains labor-intensive and prone to human error, often failing to detect subtle dishonest behaviors such as unauthorized interactions. Existing surveillance methods lack the precision required for modern exam settings, where detecting malpractice is increasingly complex. To address the challenges, AI-driven techniques leveraging machine learning models like YOLOv8 and RetinaNet are explored for real-time detection of suspicious behaviors during offline examinations. The detection systems analyze actions such as head-rotation, note-passing, and unauthorized communication, enhancing accuracy while reducing reliance on human oversight. Training the models on datasets incorporating varied environmental conditions ensures robust detection and classification of unethical behaviors. The comparative analysis demonstrates that YOLOv8 outperforms RetinaNet in controlled environments with a precision of 0.96, whereas RetinaNet excels in dynamic environments with a precision of 0.91, offering higher precision, in their respective domains. The outcomes of the paper help to strengthen proctoring systems, minimize human biases, errors, and reinforce academic integrity by ensuring a fairer conduct of examinations.
The article examines the importance of using mathematical modelling software (MMS) in higher education as an effective tool for digital transformation of the educational environment. The objective of this research is to determine the conceptual provisions regarding the optimal integration of mathematical modelling software to ensure the efficiency of higher education and prepare future specialists for the challenges of the modern world, as well as to develop recommendations for the optimal use of these tools in the educational process. The survey of modern research, regarding the use of mathematical modelling software in education, including the main trends, achievements and challenges, was conducted. The impact of these software tools on the activity of students in the educational process and their general scientific potential was investigated (on the base of the Department of Computer Sciences of the Faculty of Physics and Mathematics of the Kamianets-Podilskyi Ivan Ohiienko National University). It was determined that software tools of mathematical modelling enable to adapt the educational process to the needs of each student, providing an opportunity for individual study of the material and development of student’s interests. The use of mathematical modelling software in education is an effective tool for the digital transformation of the educational environment and the preparation of young people for the challenges of the modern world.
The world is said to shape itself by megatrends. Yet, there is widespread uncertainty about the impending transformation expected to be wrought by Artificial Intelligence (AI). A transformation expected to be without a precedent, and orders of magnitude different from those in the past. Higher Education (HE) must play a key role in a country’s adaptation to the imminent transformation. It must provide the meta-skills for the young to adapt and thrive in the new future. Meta-skills are not new – but the self-management, social intelligence, and innovation skills they encompass must be reinvented. Going forward, they are skills about skills for deploying human intelligence in conjunction with AI, effectively, efficiently, and ethically. The meta-skills can be delivered through the agency of AI, and traditional human instructors, tutors, oneself, peers, and family. These agents must be integrated to deliver the curriculum and strategy for meta-skill development, assessment, and feedback to develop the skills in the spectrum of major, minor, ability enhancement, value addition, multidisciplinary, and skill development courses of an undergraduate degree. The acquisition of meta-skills must be induced through different agent combinations through the warp and weft of pedagogy and courses. We propose a ‘Google Map’ of the challenge as a framework to chart the pathways for meta-skills development in the undergraduate curriculum.
Rapid urbanization and technological growth have intensified waste generation, especially e-waste, creating environmental and public health risks. Traditional manual segregation is inefficient, prompting the rise of smart waste management solutions. This review systematically analyzes 20 studies on IoT-enabled reverse vending machines (RVMs) and reward-based systems. The objective was to examine technologies used, their effectiveness, reward mechanisms, and research gaps. Findings show the use of Raspberry Pi, Arduino, and diverse sensors such as ultrasonic, IR, inductive, and load cells. Vision-based models (CNN, YOLOv3, YOLOv8, EfficientNet) achieved 76–100
As education continues to evolve, there is a growing need for socially engaging and culturally responsive learning tools. In this research, we explore the potential of a Collaborative Afrocentric E-Learning System to improve student motivation and learning outcomes. The proposed Afrocentric design principles offer tailored learning experiences that support diverse learning styles and promote collaboration among learners, educators, and communities. The system was developed using an iterative, user-centered design process. By including culturally relevant content and social features that encourage peer-based interactions, the system offers learners a welcoming and engaging learning environment. A mixed-methods pilot study was conducted to assess the system’s feasibility and acceptability among students of African descent. Students reported that they were engaged by the activities and were motivated to see their progress relative to others. Suggested improvements included adding interactive and personalized features, as well as using more vibrant colors, and offering flexible time settings for quizzes. The findings support the feasibility of the Collaborative Afrocentric E-Learning System and inform future iterations for broader implementation and evaluation.
This paper describes the concept and two instantiations of a gamified elearning course. We use badges that are attached to quizzes as the main gamification element in a course on software engineering for bachelor students in computer science. Our main goal is to support distributed practice without external pressure. This course concept has been instantiated at two universities in Germany and India. Since these countries have a very different cultural background, we compared the motivation and learning results on the two student groups. Current results show, that the motivation and learning results are quite similar despite the different cultural background.
Extending the notion of “having meaning” for characteristic phrases between pairs of languages used to be an acquis of experienced translators. As AI and Machine Learning tools are mostly rapidly developing, the parallelism gains unprecedented depth and width: users do not speak merely about some pairs but for a multitude of languages, while auto-correction mechanisms provide insight for traditional grammar and transformational syntax for the most bizarre pairs of spoken languages within a global perspective. This research uses advanced apps on phonetic transcription to provide relationship statements in the extended sphere of prosodic and musical attributes for vocalic alternations within the oral communication understanding. A fairly vigorous technical description is given for the app rendering the ergativity of this research. Therefore, the therapeutic dimension is accordingly accentuated for institutionalized rehabilitation units.
Programming education in Open and Distance Learning (ODL) contexts presents significant challenges, particularly in delivering timely, personalized support for learners who often struggle with abstract concepts and practical skills. This study introduces a two-stage AI-powered scaffolding approach—JavaTutorBot and CodeMentor-AI —designed to enhance programming mastery among undergraduate students enrolled in Object-Oriented Programming course at Open University Malaysia. Grounded in the theoretical foundations of Vygotsky’s Zone of Proximal Development and the scaffolding model, these tools aim to bridge conceptual understanding with hands-on practice. Stage 1 involves JavaTutorBot, a GPT-4-powered chatbot that engages learners through interactive, guided dialogues to reinforce core object-programming concept, namely Java class creation. Stage 2 transitions learners to CodeMentor-AI, a feedback-driven platform that provides automated, context-aware feedback on students’ coding attempts related to the targeted concept. Together, the tools support self-regulated learning, promote learner autonomy, and simulate human-like tutoring. A semester-long quantitative evaluation involving 74 students was carried out to assess Course Learning Outcome (CLO) attainment and assignment performance aligned with the learning concepts supported by the AI-driven tools. The findings revealed higher CLO attainment and strong assignment performance, with the majority of students scoring over 80
This research was centered on the identification of students’ emotions during learning activities linked to an Afrocentric collaborative e-learning system. We proposed the use of this AI-led emotion recognition system to assist students in learning new mathematical concepts. To evaluate and assess students and in what way, we compared three deep learning, Inception-v3, Plain CNN, and MobileNet-V2 techniques for the recognition of emotions. The Inception-v3 provided us with the most positive results in that it achieved an accuracy rate above 89
This study examines the impact of Generative AI (Gen-AI) tools, such as ChatGPT, on student learning, engagement, and perceptions across five undergraduate courses spanning three disciplines. Using a mixed-methods approach, we collected and analyzed both quantitative survey responses and qualitative insights over a semester. Findings reveal that student perceptions of Gen-AI’s value in enhancing learning were generally positive, with lower-division students exhibiting the most significant shifts in perception, while upper-division students were more cautious, often citing concerns about overreliance and misinformation. Despite these generally favorable attitudes, statistical analysis found no significant correlation between Gen-AI usage and final course grades beyond prior academic performance, suggesting that while AI can facilitate engagement, its effectiveness as a direct learning enhancer remains uncertain without structured instructional support. Additionally, student engagement patterns varied, with some learners actively leveraging Gen-AI for deep exploration while others engaged passively, demonstrating minimal interaction with AI-generated content. These disparities underscore the need for targeted AI literacy initiatives that guide students in critically and effectively integrating Gen-AI into their learning processes. These results highlight the importance in considering discipline-specific instructional strategies, and educators themselves being prepared to use Gen-AI tools effectively. As AI continues to shape educational and professional landscapes, ensuring both educators and students develop technical proficiency and critical evaluation skills will be crucial for effective usage of Gen-AI.
Students today frequently interact with AI through various learning and entertainment applications, impacting their perceptions and behaviors. The use of AI raises significant ethical issues, such as bias in decision-making and privacy concerns, along with changes in social interaction. This study aims to identify key factors influencing students’ ethical and social behaviors regarding AI usage, ultimately proposing solutions to enhance their awareness and conduct in an increasingly technological environment. To conduct this study, the research team conducted a survey and analysis based on 341 carefully selected and most relevant questionnaires. The study used both qualitative and quantitative methods. The impact of the frequency of AI use and other factors on the intention to use leading to ethical behavior in students was then assessed using SmartPLS 4 software. The results showed that the Frequency of AI use, along with factors such as AI Knowledge, AI Expected Value, Perception of Harm, Social Influence, and Attitude to the Intention to Use AI, affect ethical behavior in students. This study provides an overview of ethical behavior in students when using AI.
The demand for professionals in engineering remains high, yet dropout rates in this field are also significant. To address this, a student-centred transition from school to university is crucial. This paper presents the development and evaluation of an innovative teaching approach aimed at improving support for students in the early study phases. The concept focuses on better integration of mathematics and electrical engineering fundamentals, combined with a digitally supported introductory phase using learning analytics with MATLAB® Grader. Through empirical research and a case study, both technological advances and pedagogical innovations are addressed. Flexible learning paths and guided experiences are highlighted to meet the diverse needs of students. The improved introductory phase and the targeted use of MATLAB® Grader are described, aiming to promote individual learning success. Evaluation results show a correlation between exam performance and MATLAB® Grader usage, underlining its potential as a supportive tool. This paper contributes to the broader discussion on technology-supported learning and evidence-based teaching by sharing practical experiences and research insights. The approach aims to reduce dropout rates and enhance learning outcomes in engineering education through tailored, data-driven support mechanisms in the critical early stages of university study.
In today’s information society, the quality of higher education largely depends on the level of digital accessibility, which determines the possibilities of students and teachers to use efficiently digital technologies in the educational process. However, the effective use of these technologies is only possible if digital accessibility is provided for all the participants of the educational process. The objective of the research is to scientifically substantiate the theoretical and methodological fundamentals of digital accessibility provision as a key factor for enhancing higher education quality in the information society, which will contribute to expanding opportunities for students and increase the efficiency of educational process. Main structural components of digital accessibility are identified, with an emphasis on technical, cognitive and inclusive aspects that ensure equal access to educational resources regardless of physical, social or geographical barriers. The research generalizes modern pedagogical strategies of using digital technologies in higher education, based on the principles of inclusiveness, interactivity and open access to knowledge. Recommendations regarding the optimization of the digital infrastructure of higher education institutions, including the introduction of distance and blended learning, the development of digital literacy among students and teachers and overcoming digital inequality through government and institutional initiatives.
The paper presents a pilot project examining the integration of Generative Artificial Intelligence (GAI) models into a social media course at Utrecht University. The project involved an interdisciplinary collaboration between students from Utrecht University (Humanities/Media) and students from the University of Iași (Computer Science). The aim was to train students’ critical thinking in the use of GAI models particularly their potential to enhance creativity without neglecting possible biases in AI-generated social media content. Pre-project survey indicated varying levels of familiarity with GAI tools and social media among participating students. While post project evaluation reveals that most students reported increased creativity and productivity in social media content creation but they also expressed concerns about content authenticity and homogenization. Notably, many students struggled to detect subtle cultural and gender biases in AI-generated content, underscoring the importance of guided ethical reflection and feedback. The pilot project highlights the need for structured pedagogical support, feedback, and ethical reflection to prepare students for a responsible use of GAI in both cross-cultural and professional context.
This paper proposes a modular intrusion detection framework that uses advanced machine learning and features from the CICIDS 2017 dataset to accurately classify network attacks. The system integrates Wireshark for traffic capture and employs Python-based preprocessing to clean and standardize data for optimal model compatibility. A Random Forest classifier is integrated for durable multiclass classification, effectively detecting threats such as DDoS, infiltration, or web-based attacks. The architecture is conceived to offer scalability and reproducibility, supporting the offline analysis and the real-time deployment. Moreover, the synthetic data generation enhances model robustness by simulating diverse traffic patterns. Adaptive learning mechanisms allow the system to improve over time by incorporating new attack signatures. Through adaptive learning, the system continuously enhances its performance by integrating newly observed attack signatures. Beyond security, the proposed framework also contributes to learning technologies and training infrastructures by offering a reproducible pipeline that can be used for educational purposes, cybersecurity courses, and hands-on training in network analysis and machine learning. Tests demonstrate strong precision, recall, and F1-scores, especially for zero-day threat detection, confirming the framework’s robustness and efficiency in enterprise, industrial, and academic contexts.
In industrial systems where the reliability of machinery depends heavily on the condition of oils and coolants, effective fluid monitoring is essential. The conventional methods such as manual sampling, handheld measurements and such other have been widely adopted due to their low cost and operational simplicity. Do these methods meet the evolving demands of today’s smart manufacturing environments? The traditional manual methods only offer occasional updates, rely heavily on the operator’s judgment and can miss early warning signs of fluid degradation. Since, they come with drawbacks such as not support real-time monitoring, neither automated alerts nor long-term data logging and also key requirement for predictive maintenance. As a result, issues may go unnoticed until they cause downtime or equipment damage. As industries move toward digital transformation, these limitations hinder scalability and integration with modern industrial communication protocols with programmable logic controllers (PLCs). Our system critically examines the limitations of manual fluid condition monitoring and emphasizes the growing need for sensor-based, automated frameworks. Such systems enable real-time data acquisition, streamlined maintenance planning, and greater alignment with the objectives of Industry 4.0 and data-driven industrial ecosystems.
