Frontiers in Education: Digital Learning Innovations
15 hours 5 minutes ago
Digital literacy has become a foundational competence for participation in contemporary educational, social, and economic life. However, debates surrounding children's digital engagement often oscillate between concerns about online vulnerability and assumptions that expanding connectivity automatically promotes digital inclusion. This study examines how digital access, digital literacy, educational engagement, and vulnerability intersect among Basic 8 students in Ghana. The study employed a mixed-methods design involving a cross-sectional survey of 823 students drawn from 24 public schools across three districts in Ghana's Central Region, complemented by a six-week ethnographic study in three selected schools. Composite indices were developed to measure digital access, digital literacy, educational engagement, and vulnerability, all demonstrating acceptable reliability (Cronbach's α = 0.76–0.88). Correlation and regression analyses were used to examine relationships among key constructs, while qualitative data provided contextual insights into students’ digital experiences. Findings reveal a paradoxical pattern of digital inclusion. Despite limited ICT infrastructure within schools, students reported substantial engagement with digital technologies through home-based access and mobile devices. Digital access was positively associated with digital literacy (r = .174, p < .001) and educational engagement (r = .444, p < .001), while vulnerability was negatively associated with digital access (r = −0.692, p < .001), digital literacy (r = −.105, p < .01), and educational engagement (r = −0.284, p < .001). Importantly, an access–competence gap was identified among 21.6% of students, indicating that access exceeded competence. Regression analysis further demonstrated that this gap was the strongest predictor of vulnerability (B = 0.326, p < .001), suggesting that exposure to digital environments without corresponding competencies increases susceptibility to online risks. The findings challenge access-centred approaches to digital inclusion by demonstrating that connectivity alone does not guarantee safe, meaningful, or empowering digital participation. The study argues for competence-centred digital inclusion policies that integrate digital literacy, online safety, information verification, and critical media education within basic education systems. By operationalising the concepts of vulnerability and the access–competence gap, the study advances understanding of digital inequality in Ghana and comparable low-resource contexts across sub-Saharan Africa.
Hayford Mensah Ayerakwa
15 hours 5 minutes ago
Virtual Reality (VR) being an promising tool for immersive learning nowadays, nevertheless, in K-12 classrooms across the UAE, it's still underexplored thus highlighting a research gap in VR in the UAE context. This study collected data from twenty-five teachers in a single UAE private school who used VR with their students. Using qualitative focus group interviews and data thematic analysis, the study explores the teachers' hands-on experiences, perceived educational impacts, challenges, and practical considerations regarding VR integration in the UAE educational context. The findings discussed the discrepancies of VR's perceived educational benefits and the issues associated with integrating VR into education. Themes identified include student involvement and engagement in VR learning, lack of cultural-related curriculum for Arabic implementation, strategies used by teachers to create innovative solutions for classroom delivery with the lack of resources, technical barriers for both students and teachers, and safety-related issues concerning young children using VR technology. One point that emerged was the difference in what VR was presumed to accomplish regarding teaching and learning and how effective VR is when being actively implemented in an actual educational environment. This study highlights the need for Emirati localized content development, robust infrastructure, and continuous educational support to benefit from the VR's full potential. This paper contributes to the growing body of literature on immersive learning by providing a detailed, context-specific model for VR adoption in a non-Western educational system, offering valuable understandings for educators, policymakers, and technology developers.
Rana Zein
15 hours 5 minutes ago
IntroductionThe growing availability of generative artificial intelligence (AI) tools has created new opportunities to enhance language learning, particularly speaking practice. However, studies on the effectiveness of AI voice-chat applications in enhancing English as a Foreign Language (EFL) learners’ speaking skills and autonomy remain limited.MethodsThis mixed-methods quasi-experimental study investigated the effects of Microsoft 365 Copilot voice chat on speaking fluency, interactional competence, and learner autonomy of 52 university English as a Foreign Language (EFL) learners over eight weeks. The experimental group (n = 26) engaged in voice-based speaking tasks with Copilot, while the control group (n = 26) completed parallel peer-to-peer speaking activities.ResultsQuantitative analyses revealed that the experimental group demonstrated greater improvements in speech rate, mean length of run, and pause reduction, as well as larger improvements in turn-taking, repair strategies, and topic management. Learner autonomy scores increased significantly only for the experimental group [t(25) = 4.12, p < .001, d = 0.82, mean difference = 0.90, 95% CI [0.45, 1.35]], with 71% of participants voluntarily exceeding the required practice time. Qualitative reflective responses with eight experimental participants indicated reduced speaking anxiety, perceived usefulness of immediate feedback, and increased motivation.DiscussionThese findings suggest that AI voice chat, when integrated into learners’ existing academic tools, may support multidimensional speaking development and autonomous practice in EFL contexts.
Abbas Hussein Abdelrady
1 day 16 hours ago
IntroductionThis study investigated the effectiveness of a GenAI-based intelligent learning environment, “Neuro-Adapter,” as a cognitive scaffold for developing spatial imagination skills and reducing performance gaps based on cognitive processing styles.MethodsA quasi-experimental design was used with 45 ninth-grade students (control group, n = 23; experimental group, n = 22). Data were gathered using a Cognitive Processing Style Inventory and a Pre/Post Spatial Imagination Test over a 4-week timeline.ResultsPre-test results favored holistic-visual styles (0.05 ≥ α). On the post-test, the experimental group significantly outperformed the control group with a large effect size (η² = 0.86). Within the experimental group, a non-significant post-test difference (Sig = 0.067, Cohen's d = 0.58) indicated that the environment successfully leveled the skill gap for sequential-analytical thinkers.DiscussionThe “Neuro-Adapter” functioned effectively as an operational cognitive equalizer. We recommend integrating adaptive GenAI algorithms and tailored software scaffolding strategies into spatial education curricula.
Rawan Mahmoud Abu-Seif
3 days 15 hours ago
IntroductionThis paper examines how artificial intelligence (AI) is transforming spatial analysis in Geography, generating new analytical possibilities whilst producing epistemological tensions for geographical knowledge production, particularly in the Global South.MethodsThe study employs a qualitative, interpretive design grounded in critical geography, postcolonial theory, and science and technology studies. A systematic literature review and thematic synthesis were conducted. A total of 847 records were retrieved from Scopus, Web of Science, Google Scholar, and ACM Digital Library (2015–2025), of which 52 sources met the inclusion criteria relating to geographic relevance, peer-review status, and full-text accessibility. A PRISMA-compatible screening protocol was applied.ResultsFindings reveal that AI-enhanced spatial tools, including machine-learning-based remote sensing, predictive geospatial modelling, and large language model-assisted GIS, are expanding Geography's analytical capabilities whilst simultaneously introducing epistemological challenges linked to algorithmic bias, data extractivism, and the re-inscription of colonial spatial imaginaries in computational form. The transformative potential of AI is distributed unevenly, with researchers from the Global South facing structural barriers including data poverty, inadequate infrastructure, and exclusion from foundational AI spatial dataset development, thereby deepening North–South knowledge production asymmetries.DiscussionThe paper recommends investment in sovereign spatial data infrastructure, AI literacy frameworks for geographical research, and ethical guidelines that foreground the epistemological rights of marginalised communities. It contributes to Geography by theorising the political economy of AI-enabled spatial knowledge production and arguing that critical epistemological vigilance must accompany technical adoption.
Sibonangaye Dick Nkalanga
4 days 15 hours ago
IntroductionExtended reality (XR) technologies offer promising opportunities for engineering education, yet optimal pedagogical strategies for team-based XR projects remain underexplored.MethodsThis quasi-experimental study compared cross-disciplinary and same-branch team formation in a 15-week undergraduate AR/VR engineering course. Sixty students from nine engineering programmes participated in a progressive curriculum spanning Unity fundamentals, AR development, VR interactions, and collaborative final projects using Oculus Quest 2 headsets. Cross-disciplinary teams (n = 7 teams, 32 students) were formed with members from diverse branches; same-branch teams (n = 7 teams, 28 students) comprised students from identical programmes. Outcomes assessed via pre/post knowledge tests, project rubric evaluations, and collaboration surveys.ResultsMedium-to-large advantages for cross-disciplinary teams in learning gains (Cohen's d = 0.67, p = .013) and very large effects on collaborative learning experience (d = 1.86, p < .001). Project quality trended higher for cross-disciplinary teams (d = 1.01, p = .084).DiscussionFindings suggest that intentional cross-disciplinary team formation enhances both technical skill acquisition and collaborative competencies in XR-based engineering education. All materials, anonymized data, and analysis code are openly available via GitHub for replication.
Sunny Nanade
1 week ago
IntroductionThe increasing adoption of AI-powered conversational systems in higher education has created new opportunities for scalable student support and digitally mediated engagement. However, limited research has examined how students behaviourally engage with these systems using real-world interaction data, particularly within resource-constrained institutional contexts. This study investigates patterns of student engagement within an institutional chatbot-supported student support environment using an exploratory learning analytics approach. Drawing on the Technology Acceptance Model (TAM) as an interpretive framework, it conceptualises engagement sustainability as a behavioural lens for understanding continued interaction beyond initial system use.MethodsAn exploratory descriptive learning analytics design was employed to analyse 1,495 chatbot interaction records collected from a hybrid AI-assisted student support system deployed at a South African University of Technology between July and November 2025. Behavioural interaction data were analysed using Microsoft Power BI to identify engagement patterns, interaction persistence, support-query trends, and system-classified resolution outcomes.ResultsThe findings indicate that chatbot-supported student engagement occurs within a hybrid interaction environment combining automated conversational processing with human-assisted institutional support. Student interaction patterns were highly uneven, with a small proportion of users accounting for a disproportionately large share of system activity, while many users disengaged after initial interactions. Chatbot usage was concentrated primarily within administrative support domains, and system-classified completion rates were high for structured queries. However, extended response times associated with escalated interactions reflected institutional workflow processes rather than chatbot processing performance.DiscussionThis study demonstrates the value of behavioural interaction log analysis for understanding AI-mediated student engagement in higher education. By interpreting engagement sustainability as an observable behavioural phenomenon, it extends existing applications of the Technology Acceptance Model beyond adoption intentions to actual usage behaviour. The findings further highlight the importance of considering interaction patterns, institutional context, and hybrid human–AI support models when evaluating the effectiveness and sustainability of chatbot-supported student engagement in higher education.
Nondumiso Shabangu
1 week 4 days ago
Lived experience is a powerful medium for learning, yet most educational and workforce systems lack tools to recognize complex competencies developed in homes, communities, and workplaces. This paper introduces LivedX, a socio-technical platform designed to make lived learning visible, valid, and portable across education and employment. LivedX integrates three core components: an integrative taxonomy of human capacities aligned with NACE; a human–AI annotation pipeline for coding narratives of lived experience; and a digital platform that translates narrative evidence into interpretable profiles and credential-ready artifacts. Drawing on iterative conceptual work and large-scale human coding of learning narratives, we describe how the LivedX taxonomy was developed by mapping existing frameworks, building a hierarchical codebook, and refining it through consensus-oriented annotation practice. The system was empirically evaluated on a corpus of approximately 3,000 learner narratives, achieving strong inter-rater reliability (ICC = 0.83) and improved automated competency prediction performance, including a 12% increase in F1 score and a 5% increase in AUROC relative to baseline models. We outline a workflow in which learners submit narratives, a natural language processing engine proposes competency mappings with embedded explanations, and human reviewers retain final judgment. LivedX offers a scalable model for equity-oriented recognition of non-formal learning.
Julia Mahfouz
1 week 4 days ago
Inclusive classrooms demand learning systems that accommodate diverse student cognitive profiles while supporting the development of high-level mathematical reasoning. However, empirical evidence on the integration of artificial intelligence (AI), educational robotics, and adaptive scaffolding in inclusive mathematics learning remains limited. This study aims to analyse the effects of an AI-based robotic learning environment with adaptive scaffolding on improving students’ mathematical reasoning abilities in inclusive classrooms. The study used a quasi-experimental design with a mixed-methods approach involving 47 slow learners in an inclusive junior high school. The intervention consisted of a hybrid physical-digital learning environment that combined a line-follower robot, contextual dioramas, an interactive tablet application, and an AI system for real-time assessment and tiered assistance, including problem interpretation, procedural guidance, and symbolic support. The results showed a significant increase in students’ mathematical reasoning abilities, with the average score increasing from 47.96 to 60.66 (p < 0.001) and a very large effect size (Cohen's d = 1.58). Improvements were observed in most slow-learning students, demonstrating the system's potential to support adaptive differentiation in inclusive learning. These findings suggest that integrating tangible robotics with AI-based adaptive scaffolding can serve as an effective cognitive support system to improve students’ mathematical reasoning. This research contributes to the development of an AI-based robotic learning environment model that supports deep learning and pedagogical differentiation in inclusive mathematics education.
Riawan Yudi Purwoko
2 weeks 2 days ago
With the advancement of technologies, the use of smart learning applications has been prevalent in school education, specifically in science inquiry learning. Many studies have reported the importance, significance, and learning techniques using Inquiry-based learning, but very few studies have evaluated the multi-year analyses of smart learning applications on students' evaluations along with their impact at institutional levels. Inquiry-based learning, a student-centered pedagogical approach in which students engage in questions, problems, or scenarios with minimal direct instruction, has become recognized for its significance and effectiveness in developing critical thinking, problem-solving, and independent learning skills. This paper presents a three-year multi-cohort quasi-experimental evaluation of a smart inquiry-based mobile application named MDISTIL (Mobile-Based Digital Storytelling for Inquiry Learning) implemented with different Grade-8 student cohorts between 2023 and 2025. The aim of the study is to design and evaluate the repeated implementation of a smart science inquiry-based learning application with 240 Grade-8 students across three consecutive years. Mobile Science Inquiry (MSI) framework is used to evaluate the quantitative and qualitative data using paired-sample t-test, one-way analysis of variance ANOVA, and an evaluation of a questionnaire measuring the usability of the application. Results show that the average pre-test scores were 10.75 for girls and 9.51 for boys, while the corresponding post-test scores were 12.35 and 11.74, respectively. Across the three annual cohorts, participating students demonstrated consistent improvements in post-test scores following MDISTIL-supported inquiry activities. The descriptive results indicate that both boys and girls demonstrated improvements in post-test scores following participation in the MDISTIL-supported inquiry activities. While differences in mean scores were observed between genders, the study design and available analyses do not permit firm conclusions regarding gender effects or gender-by-intervention interactions. Teacher observations suggested that students displayed varying levels of familiarity and confidence when initially handling mobile devices, while many students demonstrated strong engagement during the final presentation activities. Further, the qualitative evaluation indicates that students were satisfied with the usability of the application. Overall, this study highlights positive learning and usability patterns associated with the repeated implementation of a smart inquiry-based learning application and provides practical insights for educators and learning system designers.
Sameena Javaid
2 weeks 3 days ago
This study investigated the use of eye-tracking technology in higher education, as well as its implications for teaching and feedback, through a systematic review guided by the PRISMA 2020 guidelines (databases and registers only). Using keywords relating to eye tracking, feedback and the educational context, 2,544 records were retrieved from Web of Science-indexed journals published by Elsevier, Springer, SAGE, Taylor and Francis, Wiley and Frontiers. Following screening, 2,490 records were excluded, leaving 54 full-text reports to be assessed for eligibility. A total of 27 empirical studies conducted in higher education contexts involving undergraduate and graduate students, adult learners and university instructors met the inclusion criteria and were synthesized. Analysis of the learning environment showed that most studies were conducted in controlled laboratory settings (88.8%, n = 24), with a smaller proportion conducted in university classrooms (11.2%, n = 3). Interactive digital materials and standard/static digital materials were equally represented, each accounting for 44.4% of the reviewed studies. The findings were organized into four outcome domains: (1) cognitive processing and learning performance, (2) attention allocation and engagement, (3) affective states and (4) instructional skills and teaching quality. Overall, this review provides evidence-based insights into how eye tracking can inform the design and evaluation of feedback in face-to-face, blended and online higher education environments.
Oktay Cem Adıgüzel
2 weeks 3 days ago
IntroductionSelecting an undergraduate major is a crucial academic decision that significantly influences students' long-term learning trajectories and career outcomes. Despite policies promoting flexible curriculum design and informed major–minor combinations, students often make suboptimal choices due to limited guidance and external influences.MethodsTo address this challenge, this study proposes a LightGBM Leaf-Embedding Neural Blender (LENB) framework that integrates gradient-boosted tree representations with neural embedding layers to predict major-course selection.ResultsUsing a comprehensive dataset of 19,921 undergraduate students across multiple academic streams— Arts, Commerce, Science, Language, and Other—the proposed model achieves high predictive performance with accuracies up to 87.50% and strong cross-validation stability. The framework incorporates class balancing (SMOTE with class weighting), AdamW optimization with early stopping, and dropout-based regularization to ensure robustness and generalization. Furthermore, SHAP-based interpretability analysis provides clear insights into the factors influencing major selection, revealing that MINOR, CGPA, and Programme are the most dominant predictors, reflecting the importance of academic background and curricular structure. In contrast, socio-demographic attributes such as Gender and Religion exhibit relatively lower influence compared to academic features, indicating reduced reliance on sensitive attributes rather than complete bias neutrality.DiscussionComparative evaluation against RandomForest, XGBoost, CatBoost, LightGBM, and TabNet demonstrates that the proposed LENB framework achieves competitive performance across multiple streams and outperforms baseline models in several cases, particularly in the Arts and Commerce streams. However, in the Language stream, Random Forest achieves slightly higher accuracy, indicating that the effectiveness of hybrid models may vary depending on the underlying data structure. By integrating explainable artificial intelligence with a hybrid learning architecture, the proposed framework delivers transparent, data- driven insights that can support educators, policy-makers, and career counselors in guiding informed academic decision-making.
Emi Kalita
2 weeks 4 days ago
This paper examines whether OpenClaw, an open-source AI agent framework, can serve as the foundation for adaptive educational AI systems. We analyse OpenClaw's architecture against established requirements for adaptive learning: student modelling, curriculum delivery, difficulty adjustment, and assessment feedback. A proposed framework maps OpenClaw's core features (persistent memory, modular skills, and retrieval-augmented generation) onto educational functions including adaptive tutoring, self-regulated learning support, and data sovereignty compliance. We extend the analysis with (a) a three-year total-cost-of-ownership comparison between local graphics processing unit (GPU) deployment and frontier cloud application programming interfaces (APIs), (b) a concrete integration path with open-source Item Response Theory libraries (py-irt, mirt-Python, EduCDM) to give difficulty adjustment a psychometric foundation, (c) a multimodal accessibility layer combining open-source automatic speech recognition (ASR; Whisper) and text-to-speech (TTS; Coqui) for learners with visual or motor impairments, and (d) a Consolidated Standards of Reporting Trials (CONSORT)-style design for a single-course feasibility pilot. Resultssuggest that OpenClaw offers significant advantages in cost, customisation, and data control, while challenges remain in technical deployment complexity, psychometric maturity, and the absence of empirical validation. The paper concludes with a concrete pilot protocol and specific research directions for classroom implementation.
Simon Baradziej
3 weeks ago
Difficulties in connecting visual representations with formal deductive reasoning continue to hinder students’ success in geometry proof construction. This study examined the association between generative artificial intelligence (AI)-supported instruction and Senior High School students’ geometric reasoning and proof construction. A quasi-experimental pre-test-post-test non-equivalent control group design was employed involving 86 students from two intact classes in a public senior high school in Ghana. The experimental group received generative AI-supported instruction, while the control group received conventional instruction. Data were collected using a Geometry Reasoning and Proof Test and a perception and attitude questionnaire. Independent-samples t-tests, paired-samples t-tests, ANCOVA, and descriptive statistics were used for data analysis. The findings revealed that students exposed to generative AI-supported instruction achieved significantly higher geometric reasoning and proof construction scores than students receiving conventional instruction, even after controlling for pre-test performance. Students in the experimental group also reported favorable perceptions and attitudes toward the instructional approach. The study contributes to the growing literature on AI-supported mathematics learning by demonstrating the potential of generative AI to function as a cognitive scaffold that supports the transition from visual intuition to deductive reasoning in geometry. The findings suggest that generative AI-supported instruction may provide a useful instructional approach for enhancing students’ geometric reasoning and proof construction.
Isaac Davor
3 weeks ago
Generative Artificial Intelligence (AI) is increasingly embedded in student support systems across primary and secondary education (K-12) and higher education, offering new possibilities for personalized learning, adaptive feedback, and scalable academic assistance. At the same time, its integration raises complex ethical challenges that extend beyond technical issues to core educational values. This conceptual paper examines ethical dilemmas associated with AI-powered student support, focusing on how they appear differently across educational contexts. While concerns such as data privacy, equity and algorithmic bias, transparency, accountability, and the preservation of human relationships are shared across sectors, their implications are context sensitive. In K–12 education, ethical priorities center on developmental vulnerability, child protection, parental consent, and teacher mediation, while in higher education, attention shifts toward student autonomy, academic integrity, surveillance, and institutional accountability. Ethical frameworks often generalize across educational levels, overlooking main developmental and institutional differences. This study provides/suggests a structured, context-sensitive ethical framework to guide the responsible implementation of AI-powered student support across educational settings.
Kleopatra Nikolopoulou
3 weeks 1 day ago
IntroductionOnline learning effectiveness depends on pedagogical design rather than platform capabilities. This study evaluates the BOOOM model (Boosting Cognition, Online Collaboration, Organizing Instruction, Overseeing Metacognition, Motivating) in teacher education.MethodsA quasi-experimental pretest-posttest control group design was employed with 90 students (Group B: experimental, n = 45; Group A: control, n = 45) over 12 weeks in Google Classroom. Data were collected through achievement tests, engagement scales, self-efficacy scales, satisfaction questionnaires, and interviews.ResultsThe BOOOM model exerted a statistically significant positive effect on academic achievement, cognitive and behavioral engagement, and pedagogical self-efficacy: F(1, 87) = 28.54, p < .001, partial η² = .247. No significant differences were found for emotional engagement or platform usability.DiscussionThe effectiveness of online learning depends more on the quality of pedagogical design than on the platform itself. The BOOOM model is proposed as a practical structured approach for active learning in teacher education.
Dana Shrymbay
3 weeks 1 day ago
The rapid integration of Artificial Intelligence (AI) into higher education underscores the urgent need to assess and foster robust AI literacy among future teachers. This cross-disciplinary competence is a critical, ethical, and equity-oriented foundation that empowers educators to personalize instruction and promote inclusive learning. However, current literature exhibits a methodological gap due to the scarcity of validated Spanish psychometric instruments. Furthermore, previous studies often evaluate pre-service teachers homogeneously, overlooking specialization-specific asymmetries and hindering the development of tailored, realistic, and effective curricula. To address this limitation, this study validates the Spanish adaptation of the Meta AI Literacy Scale (MAILS) to multidimensionally assess AI literacy. The 34-item instrument—measuring application, understanding, detection, ethics, creation, self-efficacy, and self-competence—was administered to a sample of 471 students in the Faculty of Education at the University of Salamanca across Sport and Exercise Sciences (48.6%), Primary Education (34.4%), Early Childhood Education (11.9%), and minor academic tracks (5.1%), providing high ecological validity. Psychometric validation demonstrated excellent internal consistency (overall Cronbach's alpha = .903). Overall, students exhibited strengths in practical application and self-efficacy, but only moderate levels of theoretical understanding and detection. ANOVA tests revealed significant competency disparities across seven of the nine assessed dimensions. Sport Sciences students excelled in creation (M = 7.62) and application (M = 17.22) due to their frequent use of performance-monitoring technologies. Conversely, Early Childhood Education students exhibited high digital vulnerability, scoring lowest in creation (M = 5.41), application (M = 14.38), and problem-solving self-efficacy (M = 6.34). Primary Education students demonstrated the highest resistance to algorithmic persuasion (M = 9.65). Transversely, the ethics dimension showed no significant differences (F = 0.053, p = .948), reflecting a shared, highly developed deontological awareness. In conclusion, these findings highlight the need to abandon one-size-fits-all digital training in higher education. Designing tailored curricular pathways that address the specific epistemological, technical, and affective demands of each degree is essential. Only through targeted preparation can AI literacy drive equity and social justice, empowering future educators to implement AI, open educational resources, and Universal Designfor Learning frameworks safely, ethically, and inclusively.
Olga Arranz-García
3 weeks 2 days ago
Background/objectivesArtificial intelligence (AI) Chatbots have benefited students in many ways but several issues have arisen with regards to ethical aspects and students’ critical thinking skills. This study aims to provide insights to the prevalence and frequency of AI Chatbot usage among Faculty of Medicine, Universiti Kebangsaan Malaysia (FPER UKM) undergraduate students, and explore the behavioral and psychological factors associated with AI Chatbot use and dependency.MethodsThis study employed a cross-sectional design. An online survey involving 300 undergraduate students from FPER UKM consisting of medical, nursing and emergency medicine courses was conducted using questionnaires that were adapted and screened from the Bergen Facebook Addiction Scale (BFAS) and other literature. All items were rated on a five-point Likert scale with an inclusion of four open-ended questions at the end of the questionnaire for participants to give their opinion on AI chatbot usage. Two main factors contributing to AI dependency, which are academic stress and academic self-efficacy, were measured using the same scale. Data analysis of the collected responses was then done via bivariable correlation and multivariable regression using SPSS version 29.ResultsOur study included 300 students in FPER UKM which mainly consisted of Malaysian students with only 4 of them being non-Malaysian. It was found that the prevalence of AI Chatbot usage among all undergraduate courses to be 100%. Most of the respondents used AI tools for one to three hours per week with the main purpose of studying or revision. The factors most strongly associated with AI dependency score was weekly usage of more than four hours (p < 0.001, β = 0.301), followed by academic stress (p = 0.001, β = 0.188). The final multiple regression model was significantly associated with AI Chatbot dependency (p < 0.001) with R2 of 0.182.ConclusionAll undergraduate students have used AI Chatbot previously with its main benefit on studying and revision. AI Chatbot dependency is significantly associated with duration of AI Chatbot usage and level of academic stress. Further studies are warranted to classify dependency on AI Chatbot and explore other potential correlates related to over-reliance on AI Chatbot.
Jonathan Wei De Tan
3 weeks 4 days ago
IntroductionIndonesia's extraordinary cultural diversity, which encompasses more than 1,340 ethnic groups and 718 regional languages, presents both opportunities and challenges for early childhood education. Although multicultural education is recognized as essential during the formative years, the integration of artificial intelligence (AI) models with culturally responsive pedagogy remains largely unexplored, particularly in early childhood education. This study investigates the structural relationships among teachers’ multicultural competencies (TMC), AI technology readiness (ATR), institutional support systems (ISS), the implementation of AI-based multicultural learning media (AMLMI), and the effectiveness of multicultural education management (MEME).MethodA cross-sectional survey was conducted among 500 kindergarten teachers in Central Java Province, Indonesia, using stratified random sampling. Data were collected using a structured questionnaire comprising 48 items, rated on a five-point Likert scale. The proposed structural model was analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM) with SmartPLS 4.0, including bootstrapping with 5,000 subsamples for hypothesis testing.ResultsThe measurement model demonstrated adequate convergent validity (all external loadings >.70; AVEs > .50) and strong construct reliability (Cronbach's Alpha > 0.88; CR > 0.91). The structural model showed that ISS was the strongest predictor of AMLMI (β = .571, p < .001), followed by TMC (β = .219, p < .001) and ATR (β = .163, p = .001), which collectively explain 77.1% of the variance. AMLMI, in turn, is strongly associated with MEME (β = .832, p < .001; R² = .692). Full mediation was confirmed for all indirect paths (VAF=100%), with ISS providing the largest indirect effect on MEME through AMLMI (β = .475, p < .001).DiscussionThe findings indicate that institutional support is the strongest predictor of AI-based multicultural education outcomes, while teachers’ multicultural competence and AI technology readiness serve as significant complementary predictors.The pattern of full mediation suggests that these factors must be implemented through the effective application of AI-based media to enhance multicultural education management. This study contributes an integrated empirical model that bridges the fragmented literature on multicultural pedagogy, AI readiness, and institutional support, offering actionable insights for policymakers and practitioners in culturally diverse early childhood education.
Luluk Elyana
4 weeks ago
Despite the rapid digital transformation in the UAE, research often relies on quantitative metrics that fail to capture the lived experiences of students. A significant gap exists in understanding how specific cultural and academic pressures in the UAE intersect with social media usage to influence mental well-being beyond surface-level usage statistics. This study explores how social media usage influences the mental well-being and educational engagement of university students in the UAE, highlighting the tension between social support and psychological stress. Using a qualitative, interpretivist framework, we conducted semi-structured interviews with 15 single university students between the ages of 18 and 24 who resided in the UAE, attended a UAE educational institution, and had used social media for at least six months. We selected 15 participants for this study because it allowed us to develop themes that the data strongly supported across their lived human experiences. We limited participation to single individuals to isolate the impact of social media from variables related to marital or domestic responsibilities. We selected this age range because some social media platforms only permit users over 18 to create accounts, and individuals older than 24 are generally no longer university students because they have completed their degrees. We analyzed the data using Braun and Clarke's thematic analysis approach. The analysis moved beyond descriptive listing to reveal a digital paradox categorized into four themes. Findings revealed that social media functions as a vital tool for educational support and professional networking (e.g., LinkedIn and WhatsApp groups) while simultaneously acting as a psychological stress through curated perfectionism and social comparison. Notably, findings also suggest that students are not passive victims of algorithms but active agents who use content curation as coping mechanisms to overcome digital distractions and maintain emotional regulation. This study contributes to digital well-being literature by providing a contextualized concept of intentional engagement within the UAE's unique cultural landscape. Practically, it offers evidence for the need of designing targeted interventions that promote digital literacy and healthier social media boundaries. While limited by its English-speaking sample, yet this study clearly highlights the strategies to promote healthier social media engagement in diverse linguistic and cultural context.
Seemal Ali