Electronic Journal of e-Learning

Self-efficacy as a Mediator Between AI-dialogic Scaffolding, Language Anxiety, and Speaking Confidence in Saudi EFL Context

Shadi Majed Alshraah — 2026-07-07

Obstacles remain in the form of inconsistent outcomes of AI applications in English as a foreign language (EFL) speaking instruction, particularly in Saudi contexts, where language anxiety and a sense of insecurity prevent learners from becoming more empowered through technological exposure. In this study, a mediator variable, self-efficacy, was postulated in the interaction between AI dialogic scaffolding, language anxiety, and speaking confidence. The current research employed a quantitative cross-sectional design, with data analyzed using partial least squares structural equation modelling (PLS-SEM) among 243 Saudi students studying at EFL universities. The findings established that AI-dialogic scaffolding had positive effects on speaking confidence and self-efficacy, and negative effects on language anxiety were very high. These relationships were partially mediated by self-efficacy, which is a critical psychological mediator. The results present a new model that incorporates technological and affective factors, offering meaningful theoretical and practical implications for the creation of AI-based language-learning contexts that facilitate psychological stability and skills acquisition. The originality of this research lies in empirically verifying complex mediating pathways in the context of Saudi EFL and extending the models of direct effects that most other researchers have previously explored. This study offers a rational framework that combines technological, cognitive, and affective features, thus contributing to the theoretical knowledge of both applied linguistics and educational technology. It extends beyond the examination of immediate impacts and shapes models of how scaffolding procedures align with the wavy paths through which they exert their influence. The studies suggest using an evidence-based approach in the Saudi context, and researchers should focus on enhancing self-efficacy to break anxiety and lack of self-confidence. Finally, the study illuminates that the true potential of AI in ed-tech is not just its ability to copy an interaction; rather, its capacity to be organized in a way that instills psychological strength and confidence in the messages it delivers. Thus, the study provides a distinct path for the evolution of a better, more holistic, and learner-focused digital language-learning environment.

Generative AI in University Mathematics: Attitudes and Academic-Leisure Use Patterns

Hassan Hossein-Mohand — 2026-07-07

The rapid spread of generative artificial intelligence has reshaped university e-learning, yet its incorporation into mathematics learning remains constrained by disciplinary demands for precision, justification, and epistemic scrutiny. This study examined the relationship between socio-demographic variables, access conditions, and attitudes towards AI in mathematics, and analyzed how these factors were associated with reported use of AI tools in two differentiated contexts: academic learning and leisure. A quantitative, observational, non-experimental, cross-sectional study was conducted with 869 students from the University of Granada across the Melilla, Ceuta, and Granada campuses. Data were collected through an online questionnaire that included the IAMAT scale and two dichotomous indicators of AI use. Inferential analyses were estimated with between 834 and 846 complete cases, depending on the procedure. The IAMAT scale showed adequate internal consistency (α = .796), high sampling adequacy (KMO = .888), and an interpretable two-factor structure that distinguished between usefulness/confidence and uncertainty/errors. AI use was more frequent in mathematics learning (62.6%) than in leisure (28.3%). Logistic regression models indicated that positive attitudes towards AI in mathematics increased the likelihood of use in both contexts. In academic use, older age, lack of Wi-Fi access, and membership of the Melilla campus were associated with a lower probability of use, whereas participation in voluntary work was associated with a higher probability. In leisure use, women showed a lower probability of use than men. In addition, K-means clustering identified six differentiated profiles defined by age, perceived usefulness, and distrust. These profiles discriminated academic use significantly, but not leisure use. The findings suggest that the adoption of AI in mathematics cannot be reduced to technological access alone, because it is also shaped by domain-specific cognitive and affective dispositions. From an e-learning perspective, these findings contribute to AI-supported mathematics education by showing that digital learning environments should move beyond mere access to GenAI and embed critical AI literacy, mathematical verification criteria, and reasoning-oriented tasks in which students explain, check, and revise AI-generated solutions rather than delegate the full intellectual workload to the tool.

Pedagogical Alignment, Adaptivity, and Analytics in Intelligent Tutoring Systems: A Systematic Review

Ilyass Houssam — 2026-07-16

Intelligent tutoring systems have changed quickly since large language models became widely available in 2023, raising a practical question for e-learning research: when a tutor is built on a general-purpose language model rather than on hand-encoded rules, what happens to its pedagogical grounding, learner adaptivity, and reproducibility infrastructure? Earlier reviews have examined these matters separately, but none has considered how they come together in the systems now being built. This systematic review addresses that gap through three analytical axes: pedagogical alignment with an instructional theory, adaptivity and learner modelling, and analytics standards supporting comparison and reproducibility. The review followed the SPAR-4-SLR protocol and used PRISMA 2020 as a reporting framework where the recoverable record allowed. Candidate records were screened, verified against their source documents, and coded conservatively against pre-specified criteria. Twelve primary studies were retained after source verification. Pedagogical alignment and adaptivity were each addressed by eleven of the twelve studies (92 percent), although implementation varied widely, from explicit frameworks such as Cognitive Apprenticeship, Productive Failure, and Socratic questioning to looser prompt-based behaviour. Analytics standards and reproducibility were the weakest axis, addressed by three studies (25 percent). No study combined all three axes with explicit uncertainty quantification or calibration of tutoring decisions. Controlled learning-outcome evidence was also scarce: only one study reported a controlled comparison, and its findings were preliminary. For e-learning practice, the review gives teachers, platform designers, and institutional decision-makers a realistic basis for expectation: current language-model tutoring systems are promising and often pedagogically plausible, but their learning benefits remain under-demonstrated and should be evaluated locally before large-scale adoption. For e-learning knowledge, the review contributes a source-verified synthesis of recent intelligent tutoring research and identifies a clear methodological gap: the field is progressing faster in system design than in shared evaluation, interoperable analytics, and calibrated decision-making. The paper concludes that future work should combine instructional theory, inspectable learner models, standardised logging, open benchmarks, and uncertainty-aware evaluation within the same tutoring systems.

Modeling AI Tool Adoption in Higher Education: The Role of Authentic Learning and Prompting Competence

Sultan Hammad Alshammari — 2026-07-27

As AI tools are increasingly used in higher education, understanding the factors affecting students’ adoption intentions has become theoretically and practically important. Previous studies have mainly relied on traditional technology acceptance constructs, while comparatively little attention has been given to pedagogical and competence-based conditions shaping students’ cognitive evaluations of AI systems. To address this gap, the current study builds on the technology acceptance model (TAM) by proposing authentic learning (AL) and prompt engineering competence (PEC) as precursors of perceived usefulness (PU), perceived ease of use (PEU), and behavioral intention (BI) to use AI tools. The study was based on data collected from 309 undergraduate students at the University of Ha’il. A two-step structural equation modeling (SEM) approach was employed using AMOS software. Confirmatory factor analysis confirmed construct reliability, convergent validity, and discriminant validity. SEM was then conducted to test the hypotheses. The findings show that AL significantly predicts both PU and PEU, whereas PEC significantly predicts PEU but not PU. Both PU and PEU were found to be important predictors of BI. Bootstrapping results reveal that AL affects BI through PU and PEU, while PEC affects BI entirely through PEU. The results also confirm considerable explanatory power, with an R² of .71 for BI. These findings extend TAM by reconceptualizing AL as a foundational pedagogical precursor influencing AI adoption and by clarifying the unique role of PEC in improving PEU. Integrating pedagogical and competence-based determinants into AI-enabled higher education advances technology acceptance theory and explains the AI adoption mechanism more precisely. The findings provide practical guidance for educators and instructional designers by emphasizing the importance of integrating authentic learning tasks and developing students’ prompt engineering skills to enhance meaningful AI-supported learning.

Validating an Instrument of Adult Learning Needs in Online and Blended Learning

Anh Nguyet Diep — 2026-07-31

It is a challenge to promptly respond to the learning needs of students, particularly adult learners characterized by rich lived-experiences, (sometimes) negative prior learning experience, multiple and contrasting motivation types alongside considerable socio-demographic heterogeneity. Furthermore, developing a typology of adults’ learning needs remains a challenging endeavour given the different perspectives in learning needs conceptualization and the different theories concerning adult learning in both traditional and online and blended learning (OBL). In particular, no validated instrument currently exists to identify and measure adult learners' needs in OBL contexts, representing a gap that needed to be addressed. This paper developed and validated an instrument to identify adults’ learning needs based on adult learning, self-regulated and motivational theories, and technological acceptance models. We deconstructed adults’ learning needs into three categories based on the existence, relatedness, and growth (ERG) theory of human needs. Data (N=209) were collected from adult learners following blended learning programs in eight centres of adult education in Flanders, Belgium. Confirmatory factor analysis (CFA) revealed that the instrument comprising of nine dimensions of learning needs displayed adequate model fit and factor loadings. Multivariate analysis of covariance (MANCOVA) highlighted that learners who differed in educational attainment, age, and ICT skills reported differences in their needs across the sub-dimensions. Motivational orientations, however, did not significantly differentiate learning needs at the multivariate level. The validated instrument, hence, offered practitioners a diagnostic tool for needs-based OBL design. Theoretically, the study challenged the assumption that motivational orientation alone predicted differentiated need profiles in OBL, opening new avenues for debate on the interplay between motivation, prior experience, and adult learning needs.

Multi-level Emergency Remote Teaching Framework for K–12 Distance Education: A Stakeholder-based Qualitative Study

Bünyami Kayalı — 2026-08-03

This study develops a multi-level Emergency Remote Teaching (ERT) framework for K–12 distance education, grounded in the experiences of students, teachers, parents, and administrators during the COVID-19 pandemic. Prior ERT research has predominantly focused on single stakeholder groups or produced descriptive findings without synthesizing perspectives into a structured response model. Using a qualitative case study design, semi-structured interviews were conducted with 100 participants (25 per stakeholder group) selected through maximum variation sampling. Thematic content analysis was employed to identify convergent and divergent priorities across groups, which were subsequently mapped onto the five layers of the Holistic Analytical Layer Framework (HALF): policy, system, program, course, and activity, based on their functional relevance to each layer. Findings indicate that digital infrastructure functions as a foundational system-level prerequisite shared across all stakeholder groups, while pedagogical, psychosocial, and governance priorities diverge by role. Teachers emphasized assessment reliability and instructional continuity; students highlighted engagement and psychosocial support; parents stressed communication and guidance mechanisms; while administrators prioritized governance flexibility and decision-making autonomy. The resulting framework positions ERT not as an isolated emergency improvisation but as a coordinated, multi-layered response architecture that aligns stakeholder expectations across interconnected educational layers and supports the coordination of policy, infrastructure, programmatic planning, and instructional practices during crisis periods. The study contributes to the distance education literature by transforming multi-stakeholder perspectives into an integrated framework and extending the application of HALF to emergency education contexts. For e-learning practice, the framework offers school administrators and policymakers a concrete diagnostic tool for auditing crisis-preparedness across governance, infrastructure, and instructional layers, while offering teacher educators a basis for designing professional development that addresses both technical and psychosocial readiness for remote instruction. Given its single-region, qualitative design, the framework should be understood as a context-sensitive foundation for future empirical validation rather than a universally transferable model.