Purpose: This research systematically examines how AI technologies in higher education contribute to achieving Sustainable Development Goals (SDGs), focusing on SDG 4 (quality education) and SDG 10 (reduced inequalities). Design/methodology/approach: We included peer-reviewed articles published in English between 2015 and 2025, identified through ERIC, Web of Science, Education Source Ultimate, Sustainability Reference Center and Academic Search Ultimate databases. Out of the 1,673 records, 1,360 were screened after 313 duplicates were removed. We conducted full-text reviews on 75 eligible articles and 51 met the inclusion criteria. Findings: The findings indicate that AI advances quality education through personalized learning, predictive analytics and adaptive systems, while promoting reduced inequalities via accessibility and inclusive support for underrepresented groups. Key challenges include ethical concerns, infrastructure gaps and cultural resistance. Research limitations/implications: This review is limited by its focus on English-language, peer-reviewed articles published between 2015 and 2025, potentially excluding relevant studies in other languages and publication types. Practical implications: Our review highlights the need for equitable AI implementation in HEIs by promoting ethical frameworks, infrastructure investment and educator training to maximize its potential for advancing SDGs. Originality/value: This study is the first to have a broad view of the application of AI in higher education institutions (HEIs) in the effort of achieving SDGs 4 and 10. It reveals the scope, patterns and implications of AI use and synthesizes barriers impacting its integration and their influence on fulfilling SDG 4 and SDG 10 objectives.
Mental rotation is an important aspect of spatial ability. While the importance of measuring mental rotation has been explored, disputes still exist within the literature surrounding sources of item difficulty in mental rotation tests (MRTs). Furthermore, gender differences in MRT performance are often seen but not fully understood. In the current study, we analyzed sources of item difficulty in a set of spatial ability test items using the Linear Logistic Test Model (LLTM). We found that items with more cubes, color differences, and higher rotational complexity tend to be more difficult, whereas items that contain occlusion, a mirrored structure, and a homogenous configuration type tend to be easier. Next, using Differential Component Functioning (DCF) analysis, we analyzed gender differences across these different item characteristics, finding that the number of cubes and color characteristics made questions more difficult for males when compared to females. The results and implications of this study are discussed in further detail.
1 February 2022 in The African Journal of Behavioural and Scale Development Research
Item biasness analysis is an important element in assessing educational test fairness and validity. The study examined the occurrence of item biasness in Lagos State Terminal Unified Mathematics assessment for primary schools pupils. The assessment contained 40 items of multiple choice which was developed by the examination unit of Universal Basic Education Board, Lagos State using primary 5 mathematics curriculum. Primary data of scores of 2018 primary 5 Mathematics 2nd Term Unified Examination was used for this research. A sample of 640 pupils selected through multistage sampling technique was used in the study. Two research questions guided this study. Data was analysed using descriptive and inferential statistics. The results displayed closeness of the mean and standard deviation scores for examinees groups, indicating the examinees have similar ability levels. Out of the items 16 and 17 functions differentially among the examinees based on gender and age while differential item functioning cannot be estimated based on location because it was found to comprise examinee with different ability level. The results displayed a high level of DIF. Base on the findings, it can be concluded that the examination is not free from differentially functioning items and the overall fairness shows biasness in the items. Consequences and recommendations of the study were advanced.
11 December 2024 in Chinese/english Journal of Educational Measurement and Evaluation
Machine learning has become a prominent approach for prediction in many fields of endeavor, including education. Python and R have become leading software while different packages have been developed for the purpose of carrying out machine learning analysis with differences in functions in R. This paper explored the functional and empirical differences of two major machine learning packages in R to estimate metrics. This comparative study provides a comprehensive guide for leveraging machine learning packages in R, facilitating informed decision-making, and advancing the adoption of effective methodologies in data-driven research and application. Findings in this study ensure that researchers make informed decisions in their selection of packages for analysis and understand alternatives, simplifying the technicalities of algorithm for practical researchers.
This study applies the Dichotomous Many-Facet Rasch Model (D-MFRM) to evaluate the fairness and psychometric quality of the Basic Education Certificate Examination (BECE) mathematics test in a West African context. Using a balanced sample of 1,168 students across gender and school type, the analysis examined item difficulty, person ability, and potential differential item functioning (DIF). Results indicate strong model fit, with Infit and Outfit Mean Square Errors near 1.0 and standardized residuals within ±2.0. Items were well distributed across the ability spectrum, targeting average and lower-ability students. DIF analyses revealed minimal bias across gender and school type, confirming measurement invariance. While item reliability was high (0.97), student separation reliability was low, suggesting limited differentiation among individual abilities. These findings validate the BECE mathematics test as an equitable assessment tool and highlight D-MFRM’s utility for ensuring fairness in large-scale educational testing.
11 December 2024 in Chinese/english Journal of Educational Measurement and Evaluation
机器学习已经成为包括教育领域在内的诸多领域中不可或缺的用于预测的方法。Python 和 R 是机器学习的主流编程工具。为了更好地进行机器学习分析,研究人员开发了功能各异的 R 包。本文探讨了 R 语言中两个主要的机器学习程序包在估计指标方面的功能性与经验性差异。这项比较研究为研究者高效使用R 包提供了综合指导,既能促进科学决策,又能帮助研究者在数据驱动型的研究和实践应用中采用有效的方法。本研究的发现确保研究人员在选择 R 包时做出明智决策,并理解替代方案,同时为实践者降低了算法应用的技术门槛。