This paper presents the design, deployment and evaluation of a GPT-4o-based grading and feedback pipeline for a Master’s-level Information Systems Project Management course, developed following the Design Science Research Methodology. The artefact orchestrates submission collection (Google Forms), rubric-based scoring with rubric-linked qualitative feedback (OpenAI GPT-4o via a Cloud Function) and analytics storage (BigQuery), across six consecutive case-study assignments. The AI grader supplemented rather than replaced instructor grading: official grades remained the instructor’s, while AI grades and feedback served as a second-opinion signal and additional formative feedback. The evaluation followed a pre-specified analysis plan combining a quantitative comparison of AI and instructor grades, a voluntary student survey and an automated assessment of feedback quality on the full corpus. Across 362 submissions from 66 students, the AI reproduced the instructor’s mark within the pre-specified tolerance of 0.5 points on a five-point scale in 83% of cases, with a mean absolute error (MAE) of 0.31 points and a small systematic bias of −0.16 points (95% CI [−0.19,−0.12]). Chance-corrected agreement, measured by the intraclass correlation coefficient (ICC), was moderate at best: ICC(2,1) =0.49 (95% CI [0.33,0.61]), spanning the poor-to-moderate range. Among the 112 survey responses (31% response rate), about 80% rated the feedback useful and 77% accurate. The pipeline operated at EUR 0.55 per student per assignment and delivered feedback within 24 h. The findings support using LLMs as scalable assistants for formative assessment, as a supplement to rather than a substitute for instructor evaluation.
The UNESCO declaration that teachers should be AI-competent by 2030 poses a challenge for basic school leaders, who should be at the forefront of integration within the school system. This study explored the perceptions of basic school leaders regarding AI readiness in selected Nigerian schools and the role of professional development in preparing school leaders for the effective adoption of AI in their daily instructional and administrative roles. A generic qualitative design was employed to explore school leaders’ perceptions and readiness for integrating AI into daily activities. Purposive and snowball sampling techniques were used to select participants. A thematic approach was implemented to analyse the data obtained using the Atlas. ti software. The results showed that although the participants were generally aware of AI and had positive attitudes toward it, their understanding was mostly shallow and focused primarily on functional digital tools. Most leaders learned on their own through the internet without formal training. Some of the biggest challenges to being ready included a lack of infrastructure, unreliable electricity, high data costs, and insufficient institutional support. Despite these limitations, participants were eager to learn about and utilise AI innovations. The study finds that professional development is the most crucial aspect that Nigerian school leaders need to prepare for in the context of AI. This research contributes to the field of educational leadership by demonstrating that structured, context-specific professional development, supportive infrastructure, and policy frameworks are necessary to ensure the fair and effective integration of AI in basic education.
This paper presents an application of a framework for educational social justice-oriented digital transformation, providing an analytical examination of four case studies of community-led design and implementation of digital technology with marginalized populations. The framework is rooted in a vision of social justice and reflects the principles of equity, inclusivity, participation and empowerment and ethical responsibility for marginalized populations. Through its five pillars - accessibility and infrastructure, inclusive pedagogy and curriculum, capacity building, ethical governance, and monitoring and evaluation – it is possible to see how the framework applies to the four cases. From this analysis, the framework is relevant in both analyzing previously implemented experiences of digital transformation and guiding new initiatives undertaken in practice through an educational social justice perspective.
This study examines the use of LLMs in text linguistics education, focusing on novices in generative AI (genAI) and their ability to design prompts and evaluate AI-generated outputs. Employing an action research design in higher education, it investigates prompting strategies, student perceptions, and learning outcomes in text analysis tasks. Students’ reports containing prompts and LLM outputs show that novices refine prompts through trial and error and occasionally use in-context examples or simplify complex instructions. End-of-term reflections indicate limited prompting competence but growing confidence in applying subject-specific knowledge, although students often attributed unsatisfactory results to the LLM’s limitations rather than to their own prompt formulation. Challenges such as anthropomorphizing the models and overgeneralizing limited outcomes were also observed. The findings suggest that, in this specific classroom context, students reported engaging in metacognitive reflection while using LLMs, particularly when disciplinary knowledge was already well consolidated.However, limited knowledge of prompt design remains a major obstacle.
This study explores pre-service teachers' engagement with Coding and Robotics within a teacher education programme. It aims to determine their readiness through their behavioural intention to teach this new subject in South Africa. A quantitative analysis of behavioural intention predictors was conducted on data collected from 235 pre-service teachers through a structured 5-point Likert survey. The research was grounded in the Unified Theory of Acceptance and Use of Technology (UTAUT) framework, focusing specifically on three core constructs: Performance Expectancy (PE), Effort Expectancy (EE), and Social Influence (SI), and their influence on Behavioural Intention (BI). Partial Least Squares Structural Equation Modelling (PLS-SEM) was employed to support the study's focus and emerging educational context. Results indicate that Effort Expectancy (EE) had a moderate effect on behavioural intention, while Performance Expectancy (PE) and Social Influence (SI) showed weak effects. The lack of alignment between curriculum reform and teacher education policies raises concerns about the readiness of pre-service teachers to implement Coding and Robotics effectively in a classroom setting. This study contributes to the growing body of research around educational robotics and coding by highlighting its role and pre-service teachers' preparedness by examining possible factors affecting their readiness to teach the subject.
This study develops and validates a human-centered approach to measuring small-group, multidimensional engagement in a game-based Computer-Supported Collaborative Learning (CSCL) environment. Grounded in sociocultural perspectives, the proposed framework integrates multimodal data sources, including discourse, interaction logs, and linguistic indicators, to assess behavioral, collaborative, and disciplinary engagement through theory-informed human interpretation. Human-annotated engagement ratings serve as ground truth for evaluating whether supervised machine learning models can approximate multimodal interpretive judgments, establishing a pathway toward scalable assessment. Three supervised learning algorithms were compared to examine model performance across engagement dimensions, and multiple regression analysis was conducted to investigate associations between quest-level average engagement and group learning outcomes. Results indicate that average collaborative and disciplinary engagement ratings positively predict group performance, whereas behavioral engagement alone does not significantly relate to learning outcomes. Qualitative analysis of group responses further illustrates these associations. Methodologically and theoretically, the study advances multimodal learning analytics by proposing a generalizable, human-centered workflow that integrates multimodal interpretation with model-assisted scalability, emphasizing the essential role of human interpretation in producing valid and contextually grounded engagement assessments in collaborative learning environments.
This paper develops a framework for social justice-oriented digital transformation in educational ecosystems for marginalized populations. Following explanations of the concepts of digital transformation, justice-oriented transformation and digital ecosystems in education, challenges faced by marginalized populations are outlined. Two main components of the framework are highlighted: five core principles – equity, inclusion, participation by engaging marginalized communities in design processes, ethical responsibility guaranteeing transparency and accountability, and digital competence fostering empowerment and agency; and five pillars - accessibility and infrastructure, inclusive pedagogy and curriculum, capacity building, ethical governance, and monitoring and evaluation.
This study aimed to compare the pedagogical and content quality of AI-generated assessment tasks with those developed by humans in English language teaching, using a sequential-explanatory mixed-methods design. To accomplish this, 6 human-developed lesson plans from American English File 3 and 6 equivalent AI-generated lesson plans were analyzed. Twenty experienced EFL teachers were invited to evaluate and rate 52 assessment tasks from these lesson plans based on a TPCK-aligned rubric. Follow-up interviews were conducted to better understand the teachers’ ratings. Quantitative results showed no statistically significant differences between AI- and human-developed tasks across eight quality criteria. More specifically, teachers preferred AI-generated grammar and vocabulary tasks, while favoring human-developed reading, listening, writing, and speaking tasks. Moreover, qualitative analysis highlighted AI’s strengths in generating tasks quickly, providing a diverse range, and managing rule-based content. The study also emphasized the importance of teacher mediation to ensure personalization and contextual relevance. Overall, the findings emphasize the need for human-AI collaboration in EFL assessment design. In other words, AI tools are useful as initial resources rather than complete replacements for human-created assessments.
Despite AI’s growing educational potential, little is known about science student teachers’ preparedness to integrate it into classroom practice. This study examined whether AI training level and institutional context were associated with student teachers’ TPACK. Using a quantitative, comparative, descriptive survey design, data were collected from final-year Bachelor of Education science students at one campus-based and one distance education university in South Africa (n = 186). Final-year students were selected because they are nearing entry into the profession, which makes their self-reported readiness to integrate AI particularly informative. Statistical analyses included descriptive statistics, exploratory factor analysis, reliability testing, nonparametric comparisons, and regression analysis. The findings revealed that while 64% of participants at the campus-based university reported strong TPACK, only 47.4% at the distance university did so. Regression results indicated that AI training was associated with self-reported TPACK only among students at the distance university. Specifically, completing a short course was associated with significantly weaker self-reported TPACK than receiving no training (p = .039), whereas no significant association was observed at the campus-based university (p = .607). Given the modest variance explained, these associations are interpreted cautiously. Pedagogical knowledge was the weakest TPACK component across both institutions. The findings suggest that uniform AI training may not produce consistent outcomes across contexts. In South Africa, where teacher education includes both distance and campus-based provision amid uneven digital access, AI training should be context-responsive, infrastructure-sensitive, and pedagogically grounded.
Learning design (LD) has a prominent role in ensuring the quality of teaching and learning. It should be based on student-centered approaches and intended learning outcomes, which are constructively aligned with teaching and learning activities (TLAs) and assessment. To support overarching change management and mutual learning related to the pedagogically sound and feasible design of courses, we investigated practices in LD. We used the data on the LDs of 554 courses developed in an innovative and free-to-use LD tool, which has already been used in more than 3,000 courses in total. We used probabilistic models and machine learning (Markov chain, pattern mining) to examine around 29,000 TLAs belonging to the selected courses, to establish some basic patterns related to LD. First, we examined how LD elements correlate and how learning outcome levels (according to Bloom’s taxonomy) relate to other features of LD. Second, we recognized patterns in the design of units of teaching and learning (which consist of TLAs), including the common composition in terms of learning types (Acquisition, Discussion, Investigation, Practice, Production, Assessment).
The growing adoption of artificial intelligence (AI) in African higher education offers opportunities for personalized learning, improved decision-making, and educational innovation. However, AI systems rely heavily on student and institutional data, raising significant privacy concerns within contexts characterized by evolving regulations, infrastructural challenges, and diverse socio-cultural norms. This study examines student and educator perceptions of data privacy in AI-enhanced learning and explores how these perceptions influence AI adoption in Sub-Saharan African higher education institutions. Guided by the ENTREQ framework and PRISMA procedures, a qualitative meta-synthesis was conducted on 18 studies published between 2020 and 2025. Data were analyzed using open, axial, and selective coding. The synthesis was informed by Privacy Calculus Theory and Contextual Integrity Theory. Findings indicate that students are primarily concerned about data security, transparency, consent, surveillance, and potential misuse of personal information. These concerns are intensified by limited digital literacy and infrastructural constraints. Educators acknowledge the pedagogical benefits of AI but emphasize governance, institutional accountability, ethical oversight, and capacity development. While both groups advocate stronger privacy protections and transparency, students focus on personal control and vulnerability, whereas educators prioritize compliance and governance. Privacy concerns reduce trust and constrain AI adoption, particularly in environments with weak regulatory and institutional safeguards. AI adoption in African higher education is shaped by privacy risk evaluation, contextual expectations regarding data use, and institutional trust. Strengthening governance frameworks, transparency, and institutional capacity is essential for responsible AI integration and sustainable adoption across the region.
Despite the proliferation of digital twin technology since its inception in the 1960s across engineering, aerospace, and business sectors, its adoption within higher education remains nascent. The transformative potential of digital twins in academic contexts is still evolving and not yet fully conceptualised by educators and administrators. This study addresses this gap through a bibliometric analysis of research trends from 2011 to 2025, drawing on 3463 articles from the Web of Science core collection. The findings delineate a global research landscape dominated by China, the United States, Australia, and the United Kingdom in both publication volume and citation impact. While no African nation ranks among the top ten most cited, the analysis identifies South Africa as the seventh most influential country and the Egyptian Knowledge Bank as the tenth most prominent affiliation, together representing the continent's sole significant contributors. This limited engagement underscores a substantial research disparity. The study concludes by advocating for strategic research collaborations between African higher education institutions and both these continental pioneers and global leaders, urging administrators to champion the integration of the digital twin paradigm to catalyse the development of Africa's digital educational ecosystem.
Peer learning is a promising strategy that encourages knowledge co-construction and reflection, while biased partner choice may hinder effective collaboration. Data-driven systems with interpretable learning analytics create new affordances for optimizing peer selection through knowledge awareness. Grounded in a technology-enhanced learning environment, this study investigates how social intimacy, knowledge proficiency gaps, and partner-selection autonomy affect peer learning in a Japanese university academic reading course. The data-driven conditions that underscored proficiency gaps in partner selection mitigated the influence of social intimacy and improved efficiency without compromising learning outcomes, although they alone were insufficient to inherently enhance engagement or learning quality. Helper-initiated pairing demonstrated relatively higher efficiency when learners could autonomously identify knowledge gaps and select partners. This study contributes a substantial implementation of computer-supported peer learning, an empirical analysis of key determinants of peer interaction, and a pedagogically aligned framework for integrating data-driven support into higher education to promote collaborative learning experiences. It also leads to future exploration of balancing these factors through contextualized design and continuously updated learner models incorporating multimodal data, thus extending applications to broader educational contexts.
The role of school leadership in digital transformation is increasingly recognised as central to creating inclusive, adaptive, and future-proof schools and education systems. However, the complexity of this task is often underestimated in policy, research, practice, and professional development. This article explores school leaders’ roles as key agents in co-designing inclusive and adaptive digital learning cultures within diverse sociocultural, technological, and educational contexts. Building on contributions from Thematic Working Group 2 (TWG2) at EDUsummIT 2025, this paper introduces a leadership framework comprising six inter-related parts: transformative and strategic leadership; capacity building; equity and inclusivity; policy and governance; leading change; and future-proofing. This framework reflects the evolving demands placed on leaders in digitally enhanced, constantly changing educational environments. Contributions include reflections on theoretical approaches, empirical studies, conceptual analyses, policy-focused pieces, and international cases. Together, they provide a synthesis of current challenges and innovative responses to leadership in the digital age, with a focus on practical tools, inclusive and adaptive strategies, and contextual variations in reframing schooling for the digital age.
In recent years, artificial intelligence (AI) has increasingly been integrated into educational contexts. This study examines the integration of ChatGPT with experiential marketing (XM) theory in music education, with a particular focus on lyric learning activities that emphasize emotional engagement. The application of ChatGPT combined with XM in lyric learning in music education remains relatively underexplored. This study addresses two research objectives: (1) to evaluate students' attitude toward use of ChatGPT in post-pandemic music education by integrating XM theory with the Technology Acceptance Model (TAM), and (2) to identify factors influencing junior high school students' acceptance and adoption of ChatGPT as a learning tool. Structural equation modeling was employed to analyze the data. The results indicate that the integrated XM-TAM model explains 53% of the variance in students' adoption attitudes. While traditional technology acceptance models primarily emphasize rational evaluations, the findings demonstrate that XM serves as a complementary framework by capturing affective and experiential dimensions. Specifically, experiential marketing and perceived enjoyment show significant positive associations with students' attitudes toward adopting ChatGPT, highlighting the central role of emotional engagement. These findings support an integrated framework for understanding student acceptance of AI-supported music education and offer implications for instructional design and future research in emotionally oriented arts education contexts.
The increasing importance of lifelong learning and professional development is linked to the growing usage of tertiary-level continuing professional development (CPD) through e-learning. However, high dropout rates remain a significant challenge both for adult learners seeking to update their skills and for institutions aiming to provide effective online education.This mapping review analyses 216 articles on dropout in e-learning courses for tertiary-level continuing professional development, published between 2007 and 2023. The review addresses three key research questions: what has been researched regarding dropout, how the research has been conducted, and what findings have emerged concerning dropout rates, causes, and interventions.The review finds that research in this field is dominated by quantitative methods and the focus is largely on MOOCs and on courses in information and communication technology (ICT), business, and STEM subjects. High dropout rates are common, though they varied widely, based on differences in how dropout was defined and measured. Time constraints, insufficient motivation, and course workload were cited more frequently as contributing factors. Persistence is associated with motivational support, high-quality course design, effective feedback, and social interaction. Interventions such as encouragement and guidance showed some success, but remain understudied.The review highlights a need for more qualitative methodologies, theoretical frameworks, and research on interventions, particularly outside the MOOC context and in professional/commercial adult education. To improve retention, practitioners should focus on flexible course structures, social features, assessment and feedback implementations, and approaches that foster motivation and self-regulation.
Learning analytics dashboards are widely used in self-paced online courses, yet many still emphasize progress and activity while giving limited attention to learners' emotional states and real-time self-regulation. This study examined whether an emotion-aware learning analytics dashboard (EALAD) could improve learners' meta-cognitive regulation, academic performance, and multidimensional engagement in a self-paced massive open online course (MOOC), compared with a standard learning analytics dashboard (LAD) and a no-dashboard condition. Using a randomized controlled design, 150 undergraduate students from three Iranian universities were assigned to an emotion-aware dashboard group (EDG), a standard dashboard group (SDG), or a control group (CG). The three-month intervention was implemented in a self-paced MOOC on instructional design, and all participants were selected from the intermediate range of the Oxford Quick Placement Test (OQPT), with scores between 31 and 40. Data were collected through the Metacognitive Awareness Inventory, an academic performance test, a multidimensional engagement questionnaire, and follow-up interviews. Quantitative data were analyzed using mixed-design and one-way ANCOVAs while controlling for pretest scores. The EDG showed the strongest gains across outcomes, including metacognitive regulation, partial i2 = 0.596; emotional engagement, partial i2 = 0.500; cognitive engagement, partial i2 = 0.463; and academic performance, partial i2 = 0.417, outperforming both the SDG and CG. The interview findings further showed that the EALAD helped learners recognize emotional states, respond to difficulties more promptly, and interpret progress more holistically. Participants viewed the standard dashboard as useful for pacing and progress tracking, but less helpful for understanding engagement quality. The findings suggest that integrating real-time emotional indicators with actionable dashboard prompts can support self-regulatory decision-making, meaningful engagement, and achievement in self-paced MOOC environments.
Many novice and pre-service teachers turn to social media platforms like TikTok for on-demand support. Hashtags such as #TeacherTok have become informal professional learning networks that provide accessible, peer-generated content focused on classroom strategies. Drawing on approximately 1400 TikTok videos, this study combined topic modeling with virality metrics to investigate the prevalence, themes, and engagement of classroom management content tagged with #TeacherTok. Findings identified seven salient themes with varied reach and interaction. Viral content often emphasized quick, attention-grabbing strategies; while these posts offered useful tips, they also presented the potential risk of amplifying oversimplified or problematic material due to algorithm-driven preferences. Moreover, TikTok’s curation system may build community and provide positive reinforcement, but findings also revealed gaps where less viral yet critical topics remained underrepresented. This study demonstrates the value of combining computational methods with critical media perspectives to examine large-scale teacher discourse in digital spaces.
Digital competence is a critical skill for teachers in the modern era, as technology reshapes the education eco-system. The ability to integrate digital tools effectively into teaching practices not only enhances learning outcomes but also equips students with skills that are important for their daily lives. For teachers, fostering digital proficiency is fundamental to addressing the challenges of contemporary education environments, promoting inclusive learning, and addressing the evolving demands of the digital age. This article evaluates the digital competencies of K1-12 teachers in Kosovo using the DigCompEdu framework, providing information on six competency areas in the Digital Competence for Educators Famework (DigCompEdu). Using the SELFIE tool, data were collected from 441 in-service teachers across Kosovo and analyzed through normality and reliability tests. The Kolmogorov-Smirnov (KS) and Shapiro-Wilk (SW) resulted in significant deviations from normality; therefore, the analysis proceeded with non-parametric methods, primarily using the Kruskal-Wallis test for group comparisons. The reliability resulted in strong internal consistency with a Cronbach’s Alpha value of 0.901. The results reveal varied proficiency levels, with the highest scores in the “Digital Resources'' domain and the lowest in “Teaching and Learning''. Age and experience with digital tools significantly influenced certain competencies, highlighting disparities between demographic groups. By aligning DigCompEdu with the TPACK framework, this research underscores the need to bridge the “what'' (skills) with the “how'' (application) in developing digital competence for effective integration by teachers.
Environmental health education faces the challenge of connecting students with the often-overlooked organisms coexisting in their everyday environments. This study developed and validated an augmented reality (AR)-enhanced mosquito vector control curriculum grounded in integrated curiosity-flow theory. The curriculum employed a two-stage design: Session 1 created knowledge gaps to induce deprivation-type (D-type) curiosity, while Session 2 provided AR exploration tools to facilitate transformation into interest-type (I-type) curiosity and foster flow experience. Using structural equation modeling with 423 fifth-grade students, we examined relationships among mosquito-related anxiety, curiosity types, flow experience, and continuance intention. Results demonstrated that anxiety significantly predicted both curiosity types (I-type: β = .62; D-type: β = .54), which in turn predicted flow experience (R² = .52). Flow significantly influenced continuance intention (β = .53, R² = .28), with both curiosity pathways showing significant mediation effects. Findings illuminate how AR technologies provide unique affordances, real-time feedback, autonomous navigation, and achievement visualization, that support theoretical mechanisms in environmental education. This research contributes an empirically validated framework integrating curiosity and flow theories for educational technology design.
