Boban Vesin

International Journal of Human–Computer Interaction

Exploring learner engagement in e-learning environments: A predictive analytics perspective

Mikić, V., Keković, G., Mangaroska, K., Ilić, M., Kopanja, L., & Vesin, B

https://www.tandfonline.com/doi/full/10.1080/10447318.2025.2535530#abstract

During the last decade, the embracement of learner engagement in developing educational technologies has contributed to the amalgamation of favorable pedagogical practices and advanced learning tools. New opportunities for tailoring data-driven learning designs created optimal conditions for crafting personalized, interactive e-learning environments that foster successful learning outcomes. Although a plethora of metrics exist to capture engagement, there is a need for comprehensive research that incorporates both the learner’s subjective perceptions of their engagement and the objective indicators of their actual engagement. The goal of this research is twofold: first, we aim to investigate the relationships between the interaction data on student behavior in an e-learning environment and their self-reported engagement data, and second, to design a model for predicting students’ level of engagement based on the study findings. Statistical analysis was conducted using data from (n = 45) undergraduate students at the University of South-Eastern Norway who completed a one-semester programming course, to explore relationships between their engagement and behavior in the programming tutoring system. Artificial neural networks were then used to develop a prediction model for classifying students’ engagement levels, leveraging the algorithms’ adaptability to diverse input data structures and classification efficiency. The findings highlight the importance of e-learning features like coding exercises, topic-based assessments, and explanatory hints in fostering student engagement. They also demonstrate the feasibility of predicting engagement using learner activity, interaction time, and learning outcomes. The study provides insights that inform the development of future educational designs for personalized engagement detection and improved learning outcomes.

e-Learning environment, learning engagement, behavior data, questionnaire, ANN

The 10th International Symposium on Emerging Technologies for Education – SETE, Hong Kong

Can Explainable-AI with Adaptive Questioning Substitute a Textbook? A Mixed-Methods Study in CS1

Vesin, B., Robstad, A., Sadun, R. L., Mangaroska, K., Michail Giannakos

—

—

—

The 24th International Conference on Web-Based Learning – ICWL, Hong Kong

Technology for Emotions: A Review of Tools for Enabling Children to Self-Report Their Emotions in Learning

Fladmark, A., Holmeide, Marie., Zhang, F., Vesin, B., Possaghi, I., Papavlasopoulou, S.

—

—

—

Smart Learning Environments, 12(1)

Integrating multi-modal learning analytics dashboard in K-12 education: insights for enhancing orchestration and teacher decision-making

Possaghi, I., Vesin, B., Zhang, F., Sharma, K., Knudsen, C., Bjørkum, H., & Papavlasopoulou, S

https://link.springer.com/article/10.1186/s40561-025-00410-4

Technological advancements are transforming teaching methods while offering wider windows into students’ learning journeys. Multi-modal Learning Analytics Dashboards (LADs) are tools that facilitate smart classroom orchestration by aggregating and analyzing students’ responses through sensors, such as facial expressions and heart rate, for real-time insights into student engagement and emotional states. In this study, we developed an LAD for open-ended activities in K-12 settings, where orchestration is non-linear and poses challenges for standardized evaluation methods. We engaged end users (e.g., educational researchers) in the process from the early design stages and investigated the feasibility of the LAD when used in the wild. The results show how affective data support greater awareness of students’ experiences, improving teachers’ orchestration through better decision-making and agency. Roadblocks were also identified regarding data interpretability, students’ privacy, and additional teacher workload, which can limit adoption and should be carefully addressed in future implementations. Further research should investigate students’ responses more closely and further develop strategies for the responsible, explainable, and unbiased use of student affective data in real classrooms.

—

International Journal of Human–Computer Interaction, 41(12) M22

Trust in automation (TiA): simulation model, and empirical findings in supervisory control of maritime autonomous surface ships (MASS)

Poornikoo, M., Gyldensten, W., Vesin, B., & Øvergård, K. I

https://www.tandfonline.com/doi/full/10.1080/10447318.2024.2399439

Over the past three decades, Trust in Automation (TiA) has been the subject of extensive research. However, a large portion of the research takes a “static” approach to modeling trust and views trust as a linear unidirectional phenomenon. This view fails to recognize that trust is a dynamic construct that changes over time as an outcome of prolonged interaction with automation. The present study aims to address this gap and explore the nonlinear dynamic nature of trust by developing a simulation model of Trust in Automation (TiA) that can demonstrate trust evolution, deterioration, and recovery within the context of supervisory control of Maritime Autonomous Surface Ships (MASS). Employing System Dynamics (SD) approach, the model captures trust’s non-linear and reciprocal nature through dynamic feedback loops, producing behavioral patterns consistent with empirical observations of trust. The simulation results showcase the crucial role of initial trust conditions and the alignment of expectations with system performance in fostering trust and effective automation use. The study also explores the timing of system malfunctions, revealing that early faults have a greater negative impact on trust compared to later faults of the same magnitude. We tested a segment of the proposed model in an experimental study involving 30 human participants to investigate the effects of automation malfunctions on operators’ trust and behavioral responses during the supervisory control of MASS. Results not only validated the proposed model but demonstrated a significant decline in perceived reliability and trust in automation as well as the monitoring strategy after the automation malfunction.

Trust in automation (TiA), Maritime Autonomous Surface Ships (MASS), eye-tracking, system dynamics, dynamic modeling

Applied Sciences, 15(12)

Machine Learning for Chronic Kidney Disease Detection from Planar and SPECT Scintigraphy: A Scoping Review

Vrbaški, D., Vesin, B., & Mangaroska, K

https://www.mdpi.com/2076-3417/15/12/6841

Chronic kidney disease (CKD) is a progressive condition affecting over 800 million people worldwide (more than 10% of the general population) and is a major contributor to morbidity and mortality. Early detection is critical, yet current diagnostic methods (e.g., computed tomography or magnetic resonance imaging) do not focus on functional impairments, which begin long before structural damage becomes evident, limiting timely and accurate assessment. Nuclear medicine imaging, particularly planar scintigraphy and single-photon emission computed tomography (SPECT), offers a non-invasive evaluation of renal function, but its clinical use is hindered by interpretive complexity and variability. Machine learning (ML) holds promise for enhancing image analysis and supporting early CKD diagnosis. This study presents a scoping review of ML applications in CKD detection and monitoring using renal scintigraphy. Following the PRISMA framework, the literature was systematically identified and screened in two phases: one targeting ML methods applied specifically to renal scintigraphy, and another encompassing broader ML use in scintigraphic imaging. The results reveal a notable lack of studies integrating advanced ML techniques, especially deep learning, with renal scintigraphy, despite their potential. Key challenges include limited annotated datasets, inconsistent imaging protocols, and insufficient validation. This review synthesizes current trends, identifies methodological gaps, and highlights opportunities for developing reliable, interpretable ML tools to improve nuclear imaging-based diagnostics and support personalized management of CKD.

artificial intelligence; convolutional neural networks; deep learning; nuclear imaging; nuclear medicine; renal imaging

Workshop de Aplicações Práticas de Learning Analytics e Inteligência Artificial no Brasil (WAPLA) · SBC

ReguLA: A Self-Regulated Learning Component for Online Learning Systems

Ramos, S. A., Barros, A. N., de Amorim Silva, R., Mangaroska, K., Vesin, B., & Mello, R. F

—

—

—

Proceedings of the 14th Learning Analytics and Knowledge Conference

Understanding engagement through game learning analytics and design elements: Insights from a word game case study

Mangaroska, K., Larssen, K., Amundsen, A., Vesin, B., & Giannakos, M

https://dl.acm.org/doi/10.1145/3636555.3636885

Educational games have become an efficient and engaging way to enhance learning. Analytics have played a critical role in designing contemporary educational games, with most game design elements leveraging analytics produced during gameplay and learning. The presented study tackles the complex construct of engagement, which has been the central piece behind the success of educational games, by investigating the role of analytics-driven game elements on players’ engagement. To do so, we implemented a casual word game incorporating game design elements relevant to learning and conducted a within-subjects study where 39 participants played the game for two weeks. We found that the frequency of use of different game elements contributed to different dimensions of engagement. Our findings show that five of the eight game elements implemented in the word game engage players on an emotional, motivational, and cognitive level, thus emphasizing the importance of engagement as a multidimensional construct in designing educational casual games that offer highly engaging experiences.

—

Proceedings of 21st IEEE International Symposium on Biomedical Imaging (ISBI), Vol. 21

How to Evaluate Machine Learning in Medical Imaging? A Case Study on Renal Scintigraphy

Wenceslau da Silva M. F, Batista H, Barbosa G., Mangaroska K., Vesin B.

https://www.researchgate.net/publication/387180272_How_to_Evaluate_Machine_Learning_in_Medical_Imaging_A_Case_Study_on_Renal_Scintigraphy

Machine Learning (ML) techniques have gained significant attention in the Medical Imaging community in the last years, bringing new possibilities and advances to the clinical practice. Nevertheless, the utilization of ML techniques also introduced certain challenges. For example, not ensuring reproducibility and accurate evaluation of the performance of ML models, can lead to methodological shortcomings. To understand the effects of choosing an evaluation method over another, a recent study on the automatic classification on Chronic Kidney Disease (CKD) risk factor was selected. In [4], the authors proposed to use ML algorithms (e.g., Support Vector Machines, Random Forest) to support diagnosis of CKD risk factor from scintigraphy images. The proposed approach employs different feature extractors (e.g., Local Binary Patterns, Hu’s Moments) to obtain a numeric representation from multiples scintigraphy images. However, the evaluation of the ML algorithms is carried directly on the training set, which is not recommended.

—

Artificial Intelligence Review, 56(12) M21a+

Intelligent techniques in e-learning: a literature review

Ilić, M., Mikić, V., Kopanja, L., & Vesin, B

https://link.springer.com/article/10.1007/s10462-023-10508-1

Online learning has become increasingly important, having in mind the latest events, imposed isolation measures and closed schools and campuses. Consequently, teachers and students need to embrace digital tools and platforms, bridge the newly established physical gap between them, and consume education in various new ways. Although literature indicates that the development of intelligent techniques must be incorporated in e-learning systems to make them more effective, the need exists for research on how these techniques impact the whole process of online learning, and how they affect learners’ performance. This paper aims to provide comprehensive research on innovations in e-learning, and present a literature review of used intelligent techniques and explore their potential benefits. This research presents a categorization of intelligent techniques, and explores their roles in e-learning environments. By summarizing the state of the art in the area, the authors outline past research, highlight its gaps, and indicate important implications for practice. The goal is to understand better available intelligent techniques, their implementation and application in e-learning context, and their impact on improving learning in online education. Finally, the review concludes that AI-supported solutions not only can support learner and teacher, by recommending resources and grading submissions, but they can offer fully personalized learning experience.

—

2023 4th International Conference on Communications, Information, Electronic and Energy Systems (CIEES) · IEEE

Clean Genetic Algorithm Architecture for Improved Modularity and Extensibility

Janković, Z., & Vesin, B

https://www.researchgate.net/publication/377126184_Clean_Genetic_Algorithm_Architecture_for_Improved_Modularity_and_Extensibility

—

—

IEEE Transactions on Learning Technologies (M21a)

Predicting Student Performance in a Programming Tutoring System Using AI and Filtering Techniques

Ilic, M., Kekovic, G., Mikic, V., Mangaroska, K., Kopanja, L., & Vesin, B

https://ieeexplore.ieee.org/document/10608149

In recent years, there has been an increasing trend of utilizing artificial intelligence (AI) methodologies over traditional statistical methods for predicting student performance in e-learning contexts. Notably, many researchers have adopted AI techniques without conducting a comprehensive investigation into the most appropriate and accurate approach to employ. Additionally, determining the optimal input parameters for each AI technique remains a pertinent question in this domain. This study employs machine learning (ML) and artificial neural networks (ANN) to predict student grades within a programming tutoring system. The experiment involved university students whose interaction data with the e-learning system were analyzed and used for predictions. By identifying the structural relationships between the properties of the input data, this research aims to determine the most efficient AI method for accurately predicting student performance in e-learning systems. The structure of the input data in these systems is described by variables related to individual student activities, so correlations between variables were a natural starting point for further theoretical considerations. In this manner, by applying a filtering technique based on the minimum redundancy–maximum relevance (mrMR) criterion, it was shown that correlations among predictors and between predictors and the target variable play a significant role in defining the appropriate model for predicting student grades. The results showed that ANN (the Levenberg–Marquardt algorithm with Bayesian regularization) outperformed ML methods, achieving the highest prediction accuracy. The results obtained from this study can be of great importance for learning technologies engineering and AI in general.

—

2022 IEEE PES Innovative Smart Grid Technologies Conference Europe (ISGT-Europe) · IEEE

Special Day Regression Model for Short-Term Load Forecasting

Janković, Z., Ilić, S., Vesin, B., & Selakov, A

https://ieeexplore.ieee.org/document/9960317

Short-Term Load Forecasting accuracy is profoundly affected by unexpected load shapes during so-called "special days." The lack of representative data sets for these days increases forecasting error. In this paper, the authors propose a novel method for forecasting accuracy improvements during special days. The proposed model tracks historical forecasting errors and uses the deviation trend to correct the most recent forecast. Model also contains the mechanism for recognizing hours for prediction correction on special days. Model validation was performed using Serbian Transmission System Company data and showed significant improvement for special days forecast accuracy.

—

Computer Applications in Engineering Education, 30(6)

Personalisation methods in e‐learning-A literature review

Mikić, V., Ilić, M., Kopanja, L., & Vesin, B

https://onlinelibrary.wiley.com/doi/10.1002/cae.22566

Over the years, personalisation in e-learning has evolved as a promising paradigm for matching learners' specific requirements. Although many literature reviews on personalised e-learning have been conducted, there are still voids and limitations in the comprehensive literature survey regarding how personalisation might occur in e-learning environments and benefit the teaching and learning process. In addition, no design guidelines, frameworks, or tools to guide its implementation exist, so practice can vary widely. The principal purpose of this literature review is to examine the subject, provide insights into the use of personalised techniques and systems in e-learning, and investigate its impact on the engagement and success of the students. This study applies methods of precise selection criteria to determine which of the selected publications have the strongest interconnection and relevance with the topic of e-learning personalisation. It also categorises the personalisation techniques employed by each identified system and highlights new perspectives and advances in the area.

—

ACM Transactions on Computing Education (TOCE), 22(3)

Adaptive assessment and content recommendation in online programming courses: On the use of Elo-rating

Vesin, B., Mangaroska, K., Akhuseyinoglu, K., & Giannakos, M

https://dl.acm.org/doi/10.1145/3511886

Online learning systems should support students preparedness for professional practice by equipping them with the necessary skills while keeping them engaged and active. In that regard, the development of online learning systems that support students’ development and engagement with programming is a challenging process. Early career computer science professionals are required not only to understand and master numerous programming concepts but also to efficiently learn how to apply them in different contexts. A prerequisite for an effective and engaging learning process is the existence of adaptive and flexible learning environments that are beneficial for both students and teachers. Students can benefit from personalized content adapted to their individual goals, knowledge, and needs; while teachers can be relieved from the pressure to uniformly and promptly evaluate hundreds of student assignments. This study proposes and puts into practice a method for evaluating learning content difficulty and students’ knowledge proficiency utilizing a modified Elo-rating method. The proposed method effectively pairs learning content difficulty with students’ proficiency, and creates personalized recommendations based on the generated ratings. The method was implemented in a programming tutoring system and tested with interactive learning content for object oriented-programming. By collecting quantitative and qualitative data from students who used the system for one semester, the findings reveal that the proposed method can generate recommendations that are relevant to students and has the potential to assist teachers in grading students by providing a more holistic understanding of their progress over time.

—

Sustainability, 14(8)

Students’ perceptions of ils as a learning-style-identification tool in e-learning environments

Marosan, Z., Savic, N., Klasnja-Milicevic, A., Ivanovic, M., & Vesin, B

https://www.mdpi.com/2071-1050/14/8/4426

This paper presents the evaluation of the Index of Learning Styles, an assessment tool of the Felder–Silverman learning model. A few studies have previously evaluated this tool, but as far as we know, none of them considered the learners’ opinion to achieve their goals. Considering that many studies suggest continuing with the Index of Learning Styles’ evaluation, an experimental study was conducted using Protus, developed as an adaptive learning system. Analysing the concurrent validity of the Index of Learning Styles, students’ learning preferences were acquired via two different tools: the Index of Learning Styles and the subjective questionnaire. Results suggest that the Index of Learning Styles is valid for defining learning style at the beginning of the learning process, resolving the cold-start problem. We found some differences between the results of the Index of Learning Styles and subjective assessment. By enhancing the Protus user interface with new functionality, which allows a free choice of the learning style during the learning process, we overcome the observed limitations of the Index of Learning Styles. This solution could be implemented in different personalised e-learning environments, regardless of the applied assessment tool, leading to a more reliable student model.

e-learning; learning style; ILS (Index of Learning Styles); assessment tool; learning adaptive system

Frontiers in Energy Research, 10

Gab-SSDS: an AI-based similar days selection method for load forecast

Janković, Z., Vesin, B., Selakov, A., & Berntzen, L

https://www.frontiersin.org/journals/energy-research/articles/10.3389/fenrg.2022.844838/full

The important, while mostly underestimated, step in the process of short-term load forecasting–STLF is the selection of similar days. Similar days are identified based on numerous factors, such as weather, time, electricity prices, geographical conditions and consumers’ types. However, those factors influence the load differently within different circumstances and conditions. To investigate and optimise the similar days selection process, a new forecasting method, named Genetic algorithm-based–smart similar days selection method–Gab-SSDS, has been proposed. The presented approach implements the genetic algorithm selecting similar days, used as input parameters for the STLF. Unlike other load forecasting methods that use the genetic algorithm only to optimise the forecasting engine, authors suggest additional use for the input selection phase to identify the individual impact of different factors on forecasted load. Several experiments were executed to investigate the method’s effectiveness, the forecast accuracy of the proposed approach and how using the genetic algorithm for similar days selection can improve traditional forecasting based on an artificial neural network. The paper reports the experimental results, which affirm that the use of the presented method has the potential to increase the forecast accuracy of the STLF.

—