Artificial Intelligence in Educational Decision Sciences
Artificial Intelligence in Educational Decision Sciences (AIEDS) focuses on high-quality empirical, theoretical, and methodological research that examines the role of artificial intelligence in shaping, supporting, and optimizing decision-making processes within educational systems. The journal is explicitly positioned at the intersection of artificial intelligence, educational sciences, and decision sciences, emphasizing analytical rigor, theoretical grounding, and real-world relevance. The journal publishes original research articles, systematic reviews, and conceptual papers within the following scopes: AI-Based Educational Decision Systems Design and evaluation of decision support systems, predictive models, and optimization tools for instructional planning, assessment, curriculum design, and institutional decision-making. Learning Analytics and Educational Data Science Applications of learning analytics, educational data mining, big data, and explainable AI (XAI) to inform academic, managerial, and policy decisions in education. Intelligent and Adaptive Learning Technologies Intelligent tutoring systems, adaptive and personalized learning environments, recommender systems, and human–AI collaboration in learning and teaching processes. Educational Management, Leadership, and Policy Analytics AI-driven analysis for educational leadership, governance, quality assurance, resource allocation, and evidence-based policy formulation and evaluation. Ethics, Governance, and Trust in Educational AI Studies on algorithmic fairness, bias, transparency, accountability, ethical decision-making frameworks, and regulatory implications of AI use in education. Lifelong Learning and Workforce-Oriented Decisions AI applications supporting lifelong and life-course education, vocational and higher education pathways, career guidance, employability analytics, and workforce development planning. AIEDS welcomes interdisciplinary contributions that combine artificial intelligence techniques with decision science frameworks and educational perspectives, offering robust theoretical contributions and practical implications for research, practice, and policy.
Articles
10 Documents
<![CDATA[Comparison of Dataset Proportions in SVM and Random Forest Algorithms in Detecting Student Dependence on AI in Learning ]]>
<![CDATA[Sardar Faroq Ahmd Khan]]>;
<![CDATA[Pramudya Asoka Syukur]]>;
<![CDATA[Andi Baso Kaswar]]>;
<![CDATA[Marwan Ramdhany Edy]]>
<![CDATA[ Artificial Intelligence in Educational Decision Sciences Vol 1 No 1 (2026): Artificial Intelligence in Educational Decision Sciences ]]>
Publisher : <![CDATA[PT. Academic Bright Collaboration ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
DOI: 10.66053/aieds.v1i1.6
<![CDATA[Purpose – The rapid integration of artificial intelligence (AI) in education has raised concerns about excessive student dependence, potentially undermining critical thinking and learning autonomy. This study aims to identify the most effective machine learning algorithm for detecting AI dependency in learning activities and to examine the impact of training–testing data proportion on predictive performance.Methods - This study employs the CRISP-DM framework and applies two supervised classification algorithms, Random Forest and Support Vector Machine (SVM), to a synthetic dataset of 10,000 AI-assisted learning sessions. The target variable, perceived AI assistance level, was discretised into three categories (low, medium, and high). Model performance was evaluated under four dataset split scenarios (60:40, 70:30, 80:20, and 90:10) using accuracy, AUC, precision, recall, and F1-score.Findings - The results show that Random Forest consistently outperforms SVM across all dataset proportions and evaluation metrics. The highest performance was achieved by Random Forest with a 60:40 split, yielding an accuracy of 67.6% and an AUC of 80.8%. Although SVM demonstrated stable performance, it required larger training datasets and remained inferior to Random Forest.Research limitations - The use of synthetic data and limited behavioural features restricts the generalisability of the findings. The moderate accuracy indicates that AI dependency is a complex construct not fully captured by the current model. Originality - This study provides empirical evidence on the combined influence of algorithm selection and dataset proportion in detecting AI dependency, offering practical guidance for developing early-warning systems to support responsible AI use in education.]]>
<![CDATA[Computer Vision-Driven Classroom Analytics: Real-Time Attendance Verification and Student Focus Monitoring for Data-Informed Teaching Decisions ]]>
<![CDATA[Nurhikma]]>;
<![CDATA[Aril]]>;
<![CDATA[Mushaf]]>;
<![CDATA[Muh. Yusril Anam]]>
<![CDATA[ Artificial Intelligence in Educational Decision Sciences Vol 1 No 1 (2026): Artificial Intelligence in Educational Decision Sciences ]]>
Publisher : <![CDATA[PT. Academic Bright Collaboration ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
DOI: 10.66053/aieds.v1i1.7
<![CDATA[Purpose – Student attendance and learning activity monitoring are essential for ensuring instructional quality and academic accountability. However, conventional attendance methods remain inefficient, error-prone, and vulnerable to manipulation, while existing Computer Vision-based solutions often require high computational resources and focus on attendance or engagement separately. This study aims to develop an integrated, lightweight Computer Vision-based system for automatic student attendance recording and real-time focus monitoring suitable for resource-limited educational environments.Methods – This study employs a classical Computer Vision approach integrating Haar Cascade for face detection, Local Binary Patterns Histogram (LBPH) for face recognition, and rule-based eye detection for focus classification. The system automatically records attendance, tracks focus duration, and generates real-time digital reports. System performance was evaluated under controlled classroom conditions using accuracy, precision, recall, and F1-score.Findings – Experimental results demonstrate that the proposed system achieves high recognition reliability, with face detection and recognition accuracy reaching 100% in small-scale testing. The system operates efficiently with low latency and minimal computational requirements, while successfully monitoring multiple students simultaneously and generating structured attendance and focus duration reports in real time. Research limitations – The evaluation was conducted on a limited number of students under controlled conditions, which may restrict generalisability. Further testing in larger, more diverse classroom settings is required to validate system robustness.Originality – This study presents a unified and resource-efficient solution that integrates attendance validation and real-time focus monitoring within a single platform, offering practical value for schools seeking scalable and affordable learning analytics systems.]]>
<![CDATA[AI-Driven Clustering of Social Media Consumption Patterns and Daily Productivity Using K-Means and DBSCAN in Multigenerational Respondents ]]>
<![CDATA[Nurrahmah Agusnaya]]>;
<![CDATA[Putri Nirmala]]>;
<![CDATA[M. Miftach Fakhri]]>;
<![CDATA[Fadhil Zil Ikram]]>
<![CDATA[ Artificial Intelligence in Educational Decision Sciences Vol 1 No 1 (2026): Artificial Intelligence in Educational Decision Sciences ]]>
Publisher : <![CDATA[PT. Academic Bright Collaboration ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
DOI: 10.66053/aieds.v1i1.8
<![CDATA[Purpose – The rapid development of digital technology has made social media an integral part of life across generations, yet its intensive use raises growing concerns regarding its impact on daily productivity. This study aims to analyze patterns of social media consumption behavior and their relationship with productivity across age groups using a dual clustering approach based on the K-Means and DBSCAN algorithms.Methods – The study utilizes secondary data from 3,000 multigenerational respondents, processed using Orange Data Mining through stages of data selection, normalization, and unsupervised clustering. K-Means is employed to segment respondents based on proximity to cluster centroids, while DBSCAN is applied to identify density-based behavioral patterns and detect outliers representing extreme digital usage behaviors.Findings – The results indicate that K-Means effectively maps macro-level clusters primarily differentiated by age, achieving an average Silhouette score of 0.537, which reflects stable and well-separated segmentation. In contrast, DBSCAN demonstrates superior capability in identifying micro-level behavioral patterns, particularly respondents exhibiting extreme characteristics such as excessive screen time and non-productive application usage, despite yielding a lower overall Silhouette value. The comparative analysis highlights that K-Means is more suitable for demographic-based segmentation, whereas DBSCAN provides deeper insights into localized and atypical digital behavior.Research limitations – The analysis is based on a randomly sampled subset of a publicly available dataset, which may limit the generalisability of the findings across different cultural, occupational, and socioeconomic contexts. Future studies are encouraged to incorporate longitudinal data and additional behavioral variables to capture temporal dynamics and causal relationships between social media usage and productivity.Originality – This study contributes by systematically comparing centroid-based and density-based clustering approaches within a multigenerational framework to reveal both macro-demographic and micro-behavioral patterns of digital consumption. The proposed dual clustering strategy offers a novel analytical perspective for designing more adaptive and evidence-based digital literacy and productivity enhancement policies.]]>
<![CDATA[Cloud-Based Learning Analytics Platform for English Learning: Developing the Grammarlyze Mobile Application Using Firebase Realtime Database ]]>
<![CDATA[Nur Annafiah]]>;
<![CDATA[Fatur Rahman]]>;
<![CDATA[Della Fadhilatunisa]]>;
<![CDATA[Shera Afidatunisa]]>
<![CDATA[ Artificial Intelligence in Educational Decision Sciences Vol 1 No 1 (2026): Artificial Intelligence in Educational Decision Sciences ]]>
Publisher : <![CDATA[PT. Academic Bright Collaboration ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
DOI: 10.66053/aieds.v1i1.12
<![CDATA[Purpose – This study was conducted to address the growing need for a flexible and interactive English language learning platform by leveraging cloud computing technology. Specifically, it aims to develop an Android-based English learning application, Grammarlyze, and examine how Firebase can be effectively utilized to manage and store learning materials in real time, thereby improving accessibility and user experience compared to conventional learning media. Methods – The study employed the Waterfall development method, consisting of requirement analysis, system design, implementation, testing, and maintenance. The application was developed using Android Studio with Java, while Firebase Realtime Database and Firebase Storage served as the cloud backend for managing text and video learning materials. System testing was conducted using black-box testing to evaluate feature functionality.Findings – The results show that all core features of the Grammarlyze application functioned as expected. Black-box testing confirmed that 100% of the tested features, including material access, navigation, and data synchronization, were valid. Firebase enabled real-time data management, efficient storage, and seamless retrieval of learning content, contributing to a stable and responsive learning application. Research limitations – This study is limited to basic English learning materials and does not yet include automated evaluation features such as quizzes or adaptive feedback. The findings also limited to functional testing and do not measure learning outcomes quantitatively. Originality – This research provides practical evidence of Firebase implementation as a cloud platform for English learning applications, offering a scalable and efficient model that can extended in future studies to include advanced learning analytics and assessment features.]]>
<![CDATA[AI-Based Educational Decision Analytics: K-Means Clustering of University Students’ Digital Learning Readiness Using Limited and Full Attitude Schemes ]]>
<![CDATA[Annajmi Rauf]]>;
<![CDATA[Elma Nurjannah]]>;
<![CDATA[Fredy Ganda Putra]]>;
<![CDATA[Saipul Abbas]]>
<![CDATA[ Artificial Intelligence in Educational Decision Sciences Vol 1 No 1 (2026): Artificial Intelligence in Educational Decision Sciences ]]>
Publisher : <![CDATA[PT. Academic Bright Collaboration ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
DOI: 10.66053/aieds.v1i1.19
<![CDATA[Purpose – Advancements in digital learning require students to be adequately prepared both psychologically and technologically. However, students’ attitudes toward digital learning have not yet been systematically mapped using data-driven segmentation approaches. This study aims to classify university students based on similarities in their attitudes toward digital learning using the K-Means clustering algorithm and to identify the most influential dimensions distinguishing levels of digital readiness.Methods – This study employed an exploratory quantitative design using survey data collected from 469 university students. Clustering was conducted using the K-Means algorithm implemented in the Orange Data Mining application. Two variable schemes were compared: a limited scheme comprising four constructs (Psychological Traits, Growth Mindset, Learner Motivation & Engagement, and Digital Competence) and a full scheme including six constructs with the addition of Digital Readiness & Mindfulness and Student Satisfaction. Data were normalized using Min–Max normalization, and cluster quality was evaluated using the Silhouette Coefficient.Findings – Results indicate that both schemes consistently produced two optimal clusters representing students with high and low levels of digital learning readiness. The highest Silhouette Coefficient values were obtained at K = 2 for both schemes (0.335 for the limited scheme and 0.323 for the full scheme). Psychological Traits and Learner Motivation & Engagement emerged as the most significant differentiating dimensions between clusters, followed by Digital Competence.Research limitations – The findings are limited to self-reported data and a single institutional context, which may constrain generalizability. Additionally, the cross-sectional design does not capture changes in student attitudes over time.Originality – This study contributes a comparative clustering framework that integrates psychological, motivational, and technological dimensions to map digital learning readiness. The results provide a practical foundation for designing adaptive and personalized digital learning strategies based on student readiness profiles.]]>
<![CDATA[Data-Driven Obesity Classification Integrating Genetic and Lifestyle Determinants Using Naive Bayes ]]>
<![CDATA[Yusion Gandjang]]>;
<![CDATA[Amaliah Safitri K]]>;
<![CDATA[Nabila Dwi Anugra]]>;
<![CDATA[Iyang Yuyung S]]>;
<![CDATA[Akhmad Affandi]]>
<![CDATA[ Artificial Intelligence in Educational Decision Sciences Vol 1 No 2 (2026): Artificial Intelligence in Educational Decision Sciences ]]>
Publisher : <![CDATA[PT. Academic Bright Collaboration ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
DOI: 10.66053/aieds.v1i2.21
<![CDATA[Purpose – This study aims to develop a data-driven obesity classification framework that integrates genetic predisposition and lifestyle determinants using the Naive Bayes algorithm, while empirically evaluating optimal training–testing data proportions for health decision support systems.Methods – A systematic computational workflow was applied to a public obesity dataset comprising 2,112 records, which was refined to 1,259 valid instances after preprocessing. Genetic indicators and lifestyle-related variables were encoded and classified into four obesity categories: normal weight, obesity type I, obesity type II, and obesity type III. The Naive Bayes model was evaluated using three training–testing data partition ratios (75:25, 80:20, and 85:15). Model performance was assessed using six metrics: Area Under the Curve (AUC), classification accuracy, F1-score, precision, recall, and Matthews Correlation Coefficient.Findings – The results demonstrate that the 80:20 and 85:15 data partitions achieved the highest performance, with an accuracy of 0.878 and an AUC of 0.979. The model showed excellent sensitivity in identifying severe obesity cases, while moderate misclassification occurred between obesity type I and type II due to phenotypic overlap in lifestyle patterns.Research limitations – This study relies on a single public dataset and lacks population-specific genetic calibration, which may limit generalizability to diverse regional contexts.Originality – This study provides empirical validation of a probabilistic obesity classification framework that integrates genetic and lifestyle factors, offering an interpretable and computationally efficient approach to support data-driven health decision making.]]>
<![CDATA[Data-Driven Clustering of Stunting Prevention Services for Pregnant Women and Infants Using Fuzzy C-Means ]]>
<![CDATA[Hanum Zalsabilah Idham]]>;
<![CDATA[Ayu Safitri]]>;
<![CDATA[Andi Akram Nur Risal]]>;
<![CDATA[Dewi Fatmarani Surianto]]>;
<![CDATA[Firdaus]]>
<![CDATA[ Artificial Intelligence in Educational Decision Sciences Vol 1 No 2 (2026): Artificial Intelligence in Educational Decision Sciences ]]>
Publisher : <![CDATA[PT. Academic Bright Collaboration ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
DOI: 10.66053/aieds.v1i2.22
<![CDATA[Purpose – This study addresses persistently high stunting rates in South Sulawesi, Indonesia, which remain above national targets despite declining trends. We developed a clustering model to overcome limitations of traditional methods in handling complex health data with overlapping characteristics, aiming to identify priority regions requiring targeted interventions.Methods – Using 2,267 structured records from Satu Data Indonesia covering maternal and child health indicators, we implemented Fuzzy C-Means (FCM) algorithm with systematic preprocessing, optimal cluster determination via Elbow Method, and quality validation using Silhouette Coefficient.Findings – Analysis revealed three distinct clusters for pregnant women (representing good, moderate, and low service coverage areas) and three corresponding clusters for infants. Validation showed Silhouette values ranging from 0.204 to 0.645, indicating variable cluster separation quality with Cluster 0 pregnant women achieving highest cohesion (0.638) and Cluster 2 infants showing strongest separation (0.645).Research limitations – Data quality limitations affected cluster cohesion in some areas, particularly Cluster 1 infants (0.204 Silhouette value), constraining generalizability. The FCM approach accommodates real-world data complexity better than rigid clustering methods but requires high-quality input data.Originality – This research contributes an adaptive framework for evidence-based stunting prevention through sophisticated data-driven segmentation. Findings offer immediate practical value for health policymakers in resource allocation and intervention planning, with potential adaptation to other regional contexts facing similar public health challenges.]]>
<![CDATA[Predicting Student Dependency on ChatGPT for Academic Tasks Using Naive Bayes Classification ]]>
<![CDATA[Risha Febrianti]]>;
<![CDATA[Sul Fitriana]]>;
<![CDATA[Asrafah]]>;
<![CDATA[Stephen Amukune]]>
<![CDATA[ Artificial Intelligence in Educational Decision Sciences Vol 1 No 2 (2026): Artificial Intelligence in Educational Decision Sciences ]]>
Publisher : <![CDATA[PT. Academic Bright Collaboration ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
DOI: 10.66053/aieds.v1i2.23
<![CDATA[Purpose – This study aims to predict and classify the level of student dependency on ChatGPT in completing academic tasks using the Naive Bayes algorithm to support data-driven decision making in higher education.Methods – A quantitative survey approach was employed involving 254 active undergraduate students from the Department of Informatics and Computer Engineering at a public university in Indonesia. Data were collected through a Likert-scale questionnaire measuring five behavioral indicators: purpose of ChatGPT use, interaction frequency and duration, understanding of generated outputs, trust in AI responses, and learning independence. The collected data were cleaned, numerically encoded, and labeled into three dependency categories (low, medium, high). A Naive Bayes classification model was implemented using Orange Data Mining and evaluated under three data split scenarios: 90:10, 80:20, and 70:30.Findings – The results indicate that the 70:30 data split achieved the highest classification performance, with an AUC value of 0.973, accuracy of 85.3%, F1-score of 0.866, and precision of 0.909. These results demonstrate that the Naive Bayes algorithm is effective in identifying distinct patterns of student dependency on ChatGPT based on multidimensional behavioral data.Research limitations – This study is limited to a single academic program and relies on self-reported questionnaire data, which may constrain the generalizability of the findings across different educational contexts.Originality – This study provides empirical evidence on the application of probabilistic classification models to assess student dependency on generative AI, contributing to educational decision sciences by informing institutional policies on balanced and responsible AI use in higher education.]]>
<![CDATA[Clustering Multi-Indicator Learning Outcomes of Vocational High School Students: A Comparison of K-Means and DBSCAN ]]>
<![CDATA[Muhammad Fikri Aqil]]>;
<![CDATA[Irwansyah Suwahyu]]>
<![CDATA[ Artificial Intelligence in Educational Decision Sciences Vol 1 No 2 (2026): Artificial Intelligence in Educational Decision Sciences ]]>
Publisher : <![CDATA[PT. Academic Bright Collaboration ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
DOI: 10.66053/aieds.v1i2.24
<![CDATA[Purpose – This study aims to compare the performance of K-Means and DBSCAN algorithms in clustering vocational high school students’ learning outcomes in the Network Administration subject to support data-driven educational decision making.Methods – A quantitative experimental approach was employed using secondary academic data from vocational students. The variables analyzed included final examination scores, midterm examination scores, assignments, attendance, attitudes, and learning activities. Clustering was conducted using K-Means and DBSCAN algorithms implemented through data analysis software. Cluster quality and separation were evaluated using silhouette coefficients to assess the effectiveness of each algorithm in grouping student learning outcomes.Findings – The results show that K-Means produces relatively stable and interpretable clusters when student performance data exhibit more uniform distributions. In contrast, DBSCAN demonstrates stronger capability in handling noisy data and identifying students with extreme performance levels as outliers. Both algorithms successfully reveal meaningful patterns in student learning outcomes, but differ in their sensitivity to data distribution and noise.Research limitations – This study is limited to a single vocational subject and one institutional context, which may restrict the generalizability of the findings to other vocational domains.Originality – This study provides empirical evidence on the comparative performance of partition-based and density-based clustering algorithms using multi-indicator learning outcome data in vocational education.]]>
<![CDATA[Supporting Academic Library Collection Decisions Using K-Means–Based Book Recommendation ]]>
<![CDATA[Wahyuni Edsa Safira]]>;
<![CDATA[Saif Mohammed]]>
<![CDATA[ Artificial Intelligence in Educational Decision Sciences Vol 1 No 2 (2026): Artificial Intelligence in Educational Decision Sciences ]]>
Publisher : <![CDATA[PT. Academic Bright Collaboration ]]>
Show Abstract
|
Download Original
|
Original Source
|
Check in Google Scholar
|
DOI: 10.66053/aieds.v1i2.25
<![CDATA[Purpose - This study aims to develop a data-driven book recommendation system to support academic library collection management using the K-Means clustering method.Methods - The study utilized book borrowing data from the Library of the Department of Informatics and Computer Engineering at Makassar State University collected over a 22-month period. Borrowing records were grouped by book categories and monthly borrowing frequencies, then processed into numerical variables. The K-Means algorithm was applied to identify borrowing pattern clusters, and cluster quality was evaluated using the Silhouette Coefficient to assess cohesion and separation.Findings - The analysis produced three distinct clusters representing different borrowing behaviors. Programming and information technology books formed the most frequently borrowed cluster, research methodology books showed increased demand during specific academic periods, and education and learning methods books exhibited relatively lower borrowing intensity. The average Silhouette Coefficient value of 0.35 indicates a moderate yet acceptable clustering structure for recommendation and managerial purposes.Research limitations - This study is limited to historical transaction data from a single departmental library and does not incorporate user profiles or qualitative preference data, which may restrict generalizability to other academic library contexts.Originality - This study contributes empirical evidence on the use of K-Means clustering for book recommendation and decision support in academic libraries, demonstrating how borrowing pattern analysis can inform data-driven collection management and improve the relevance of library services. The findings also highlight the practical role of clustering analytics in supporting efficient resource allocation and evidence-based planning within higher education libraries and departmental level strategic decisions.]]>
