Ahlam Almalawi

Mrs. Ahlam Almalawi

A PhD student in Machine Learning .

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
13 December 2024 in Big Data and Cognitive Computing
Verified
Journal Article
Open access
PDF file
AAAhlam Almalawi
Ahlam Almalawi
Ben C. SohBen C. Soh
Ben C. Soh
ALAlice Li
Alice Li
HSHalima Samra
Halima Samra
This systematic literature review evaluates predictive models in education, focusing on their role in forecasting student performance, identifying at-risk students, and personalising learning experiences. The review compares the effectiveness of machine learning (ML) algorithms such as Support Vector Machines (SVMs), Artificial Neural Networks (ANNs), and Decision Trees with traditional statistical models, assessing their ability to manage complex educational data and improve decision-making. The search, conducted across databases including ScienceDirect, IEEE Xplore, ACM Digital Library, and Google Scholar, yielded 400 records. After screening and removing duplicates, 124 studies were included in the final review. The findings show that ML algorithms consistently outperform traditional models due to their capacity to handle large, non-linear datasets and continuously enhance predictive accuracy as new patterns emerge. These models effectively incorporate socio-economic, demographic, and academic data, making them valuable tools for improving student retention and performance. However, the review also identifies key challenges, including the risk of perpetuating biases present in historical data, issues of transparency, and the complexity of interpreting AI-driven decisions. In addition, reliance on varying data processing methods across studies reduces the generalisability of current models. Future research should focus on developing more transparent, interpretable, and equitable models while standardising data collection and incorporating non-traditional variables, such as cognitive and motivational factors. Ensuring transparency and ethical standards in handling student data is essential for fostering trust in AI-driven models.

13 May 2026 in Applied Sciences
Verified
Journal Article
Open access
PDF file
MAManal Alanazi
Manal Alanazi
ALAlice Li
Alice Li
AAAhlam Almalawi
Ahlam Almalawi
HSHalima Samra
Halima Samra
Ben C. SohBen C. Soh
Ben C. Soh
The rapid integration of artificial intelligence (AI) into higher education is reshaping teaching, learning, and assessment, particularly in programming education. While AI coding assistants can enhance feedback, scaffolding, and student engagement, their educational value depends on pedagogical alignment, institutional readiness, and faculty practice, not merely technical capability. Existing adoption frameworks, however, inadequately address these pedagogical and institutional dimensions in domain-specific contexts. This study proposes the AI Coding Assistant Adoption Framework (AICAAF), a theoretically grounded model integrating the Technology Acceptance Model (TAM), the Unified Theory of Acceptance and Use of Technology (UTAUT), and Self-Determination Theory (SDT). The framework was developed iteratively from prior literature and refined through faculty perspectives. It conceptualises adoption across four interrelated dimensions: usability, pedagogical adequacy, institutional readiness, and faculty engagement. Using PyChatAI as an instrumental case study, this qualitative research draws on semi-structured interviews with 15 faculty members teaching programming courses at Jouf University, a public institution in Saudi Arabia operating in a low- to mid-resource context. Data were analysed using reflexive thematic analysis. Findings indicate that PyChatAI is intuitive and beneficial for novice learners, particularly through instant feedback and automated error correction. However, its pedagogical value is limited in advanced and industry-aligned contexts. Institutional barriers, such as inadequate infrastructure, limited technical support, and the absence of policy frameworks, significantly constrain effective integration. Despite this, faculty expressed strong commitment to adopting AI tools, proposing strategies including curriculum redesign, professional development, and gamified instruction. The study reconceptualises AI adoption as a pedagogical and institutional transformation rather than a purely technological shift. The AICAAF provides a robust foundation to guide curriculum design, teaching practice, and policy development for responsible AI integration in programming education.

3 August 2026 in Frontiers in Education
Verified
Journal Article
Open access
PDF file
AAAhlam Almalawi
Ahlam Almalawi
BSBen Soh
Ben Soh
ALAlice Li
Alice Li
AAAfrah Alanazi
Afrah Alanazi
HSHalima Samra
Halima Samra
Introduction: Adaptive assessment has emerged as a promising approach for addressing learner heterogeneity in digital and online education, yet many existing systems rely on predefined psychometric assumptions that inadequately capture the complex, non-linear relationships among learner characteristics, engagement patterns, and academic performance. This study reports an exploratory feasibility study of a data-driven framework for personalised adaptive question recommendation, designed to estimate anticipated item difficulty using a multivariate set of learner indicators comprising prior academic achievement, engagement levels, study behaviours, and prior course experience. Because the anticipated-difficulty label is derived from four of these same learner indicators, the framework is best understood as a proof-of-concept for recovering a theoretically motivated scoring rule rather than as an independently validated measure of difficulty. Methods: To explore this proof-of-concept, an exploratory research design was adopted using an authentic dataset of 71 undergraduate students enrolled across two courses at a higher education institution in Saudi Arabia. A multilayer perceptron (MLP) model was employed to capture non-linear interactions among learner variables, with preprocessing safeguards including ordinal-preserving feature encoding, LOWESS-based outlier suppression, and stratification-controlled oversampling applied exclusively to the training partition. Model performance was evaluated using Repeated Stratified K-Fold cross-validation across nine classifiers, with holdout test evaluation and post hoc feature influence analysis conducted to support pedagogical interpretability. Results: The proposed MLP (RMSprop) achieved 93.33% test accuracy (Wilson 95% CI: 70.2%–98.8%, n = 15), demonstrating stable predictive performance and generalisation beyond training data despite sample size constraints. However, the small holdout sample, and the single observation underlying the Hard-class result in particular, mean these figures should be read as indicative rather than precise. Feature analysis indicated that learner engagement and prior academic performance were the most influential contributors to anticipated difficulty estimation, while misclassification patterns were primarily confined to adjacent difficulty levels, reflecting the inherent continuity of learner challenge. Discussion: The findings suggest that data-driven adaptive mechanisms can complement traditional assessment approaches by providing context-sensitive, learner-relative difficulty estimation aligned with formative assessment principles. Fairness was not empirically evaluated in this study and is identified as a direction for future work rather than a current contribution. While exploratory in nature, this work contributes to bridging the gap between machine learning–based modelling and principled educational assessment design, offering a foundation for future large-scale and multi-context investigations.

11 August 2026 in Frontiers in Education
Verified
Journal Article
Open access
PDF file
AAAhlam Almalawi
Ahlam Almalawi
BSBen Soh
Ben Soh
ALAlice Li
Alice Li
MAManal Alanazi
Manal Alanazi
HSHalima Samra
Halima Samra
Introduction: Online assessments often fail to differentiate student performance because scores compress near the maximum, a ceiling effect that can make a genuine engagement-performance relationship statistically invisible, whether or not one exists. This study demonstrates that risk. Methods: Among undergraduate female computer science students at King Khalid University (KKU), Saudi Arabia ( N = 242), just over half achieved perfect assessment scores, and no significant relationship between behavioural engagement and performance emerged across correlation, regression, or classification analysis. Results: Cluster analysis, which does not depend on score variance in the same way, tells a different story: it reveals four distinct engagement-performance profiles, including a high-effort, low-performance group invisible to the correlational approach. A descriptive faculty survey ( N = 20) points to a plausible cause: most assessments are custom-designed rather than drawn from difficulty-calibrated item banks and motivates a preliminary Adaptive Online Assessment Design Framework (AOADF) for future testing. Discussion: The central contribution is not a claim about whether engagement predicts performance, but a demonstration that null results in ceiling-affected assessment data reflect an instrument limitation rather than an absent relationship, with direct implications for how learning-analytics findings should be interpreted across online assessment research.