Language：Others   Publishing type：Research paper (other academic)  

Language learning applications usually estimate the learner's language knowledge over time to provide personalized practice content for each learner at the optimal timing. However, accurately predicting language knowledge or linguistic skills is much more challenging than math or science knowledge, as many language tasks involve memorization and retrieval. Learners must memorize a large number of words and meanings, which are prone to be forgotten without practice. Although a few studies consider forgetting when modeling learners' language knowledge, they tend to apply traditional models, consider only partial information about forgetting, and ignore linguistic features that may significantly influence learning and forgetting. This paper focuses on modeling and predicting learners' knowledge by considering their forgetting behavior and linguistic features in language learning. Specifically, we first explore the existence of forgetting behavior and cross-effects in real-world language learning datasets through empirical studies. Based on these, we propose a model for predicting the probability of recalling a word given a learner's practice history. The model incorporates key information related to forgetting, question formats, and semantic similarities between words using the attention mechanism. Experiments on two real-world datasets show that the proposed model improves performance compared to baselines. Moreover, the results indicate that combining multiple types of forgetting information and item format improves performance. In addition, we find that incorporating semantic features, such as word embeddings, to model similarities between words in a learner's practice history and their effects on memory also improves the model.

DOI： 10.1145/3576050.3576062

Scopus

researchmap

Language：Others   Publishing type：Research paper (scientific journal)   Publisher：Smart Learning Environments  

With the increasing use of digital learning materials in higher education, the accumulated operational log data provide a unique opportunity to analyzing student learning behaviors and their effects on student learning performance to understand how students learn with e-books. Among the students’ reading behaviors interacting with e-book systems, we find that jump-back is a frequent and informative behavior type. In this paper, we aim to understand the student’s intention for a jump-back using user learning log data on the e-book materials of a course in our university. We at first formally define the “jump-back” behaviors that can be detected from the click event stream of slide reading and then systematically study the behaviors from different perspectives on the e-book event stream data. Finally, by sampling 22 learning materials, we identify six reading activity patterns that can explain jump backs. Our analysis provides an approach to enriching the understanding of e-book learning behaviors and informs design implications for e-book systems.

DOI： 10.1186/s40561-021-00183-6

Web of Science

Scopus

PubMed

researchmap

Language：Others   Publishing type：Research paper (scientific journal)  

Recommendation systems need a deeper understanding of users and their motivations to improve recommendation quality and provide more personalized suggestions. This is especially true in the education domain, the more about the student is known, the more useful recommendations can be made. However, although many studies on the course recommendation exist, studies on the students’ course selection motivations in universities are limited. This study investigates the factors that contribute to students’ choice when selecting courses in universities to better understand student perceptions, attitudes, and needs and leverage data-driven approaches for recommending and explaining the recommendations in university environments. A qualitative interview for university students (N = 10) comprised of open-ended questions as well as a questionnaire for students (N = 81) was conducted, aiming to investigate the main reasons behind their choices. The results of this study show that students highly value the course contents and the benefits of the course towards their future careers. Furthermore, students are influenced by other reasons such as the possibility of obtaining a higher grade, the popularity of professors, and recommendations from peers. Next, we extract the main categories of students’ motivations and analyzed the questionnaire data by employing statistical analysis methods as well as the k-means clustering algorithm to identify different types of students in terms of course selection. Based on our findings, we discuss implications for designing more personalized course recommendation systems.

DOI： 10.1186/s40561-021-00177-4

Language：Others   Publishing type：Research paper (scientific journal)  

The abundance of courses available in a university often overwhelms students as they must select courses that are relevant to their academic interests and satisfy their requirements. A large number of existing studies in course recommendation systems focus on the accuracy of prediction to show students the most relevant courses with little consideration on interactivity and user perception. However, recent work has highlighted the importance of user-perceived aspects of recommendation systems, such as transparency, controllability, and user satisfaction. This paper introduces CourseQ, an interactive course recommendation system that allows students to explore courses by using a novel visual interface so as to improve transparency and user satisfaction of course recommendations. We describe the design concepts, interactions, and algorithm of the proposed system. A within-subject user study (N=32) was conducted to evaluate our system compared to a baseline interface without the proposed interactive visualization. The evaluation results show that our system improves many user-centric metrics including user acceptance and understanding of the recommendation results. Furthermore, our analysis of user interaction behaviors in the system indicates that CourseQ could help different users with their course-seeking tasks. Our results and discussions highlight the impact of visual and interactive features in course recommendation systems and inform the design of future recommendation systems for higher education.

DOI： 10.1186/s41039-021-00167-7

Language：Others   Publishing type：Research paper (other academic)  

DOI： 10.1007/978-3-032-34374-1_11

researchmap

DOI： https://doi.org/10.1080/10447318.2026.2712661

Publisher：Lecture Notes in Computer Science  

Effective learning support requires understanding not only what learners know but also how accurately they perceive their own understanding. This metacognitive dimension, known as knowledge monitoring, fundamentally influences self-regulated learning, yet this dimension remains underexplored in current systems. This paper introduces the Capture-Calibrate-Coach (3C) framework for adaptive learning support. The Capture phase extracts learners’ perceived knowledge states from open-ended self-reports to construct a heterogeneous graph linking learners and knowledge concepts. The Calibrate phase applies a heterogeneous graph neural network to infer latent perceived states for concepts not explicitly mentioned, enabling systematic knowledge monitoring assessment. The Coach phase classifies learners into five metacognitive patterns and delivers personalized feedback addressing both knowledge gaps and calibration errors. Evaluation with 684 students demonstrates 85.21% AUC in predicting latent perceived states, significantly outperforming baseline methods. A user study with 47 participants shows positive reception of feedback quality, with participants particularly valuing concrete feedback on knowledge gaps and actionable study guidance. These findings advance AI-based learning support toward metacognitive teammates that foster accurate self-awareness while supporting knowledge growth.

DOI： 10.1007/978-3-032-29755-6_2

Scopus

Authorship：Last author  

Authorship：Lead author  

Publisher：Lecture Notes in Computer Science  

This study uses a process-oriented mixed-method analysis to examine learners solving programming comprehension tasks in an online support system, integrating interaction logs, wearable eye tracking, and reflective interviews. We compare low and high performing learners’ task-solving processes and characterize their regulatory strategies. Results indicate that performance differences were more clearly reflected in process coordination patterns than in aggregate exposure measures. Low performing learners showed a material-centered workflow, whereas high performing learners more systematically integrated hints and note-taking as intermediate steps, reflected in richer attention transitions and sequences. Interviews further revealed an iterative regulation cycle from task framing to decision commitment, with concrete triggers and barriers for hint uptake. The findings point to actionable process level signals of self-regulated help seeking that can guide adaptive AI support toward improving coordination at decision bottlenecks, rather than simply increasing the frequency of resource access.

DOI： 10.1007/978-3-032-29770-9_8

Scopus

Publisher：Communications in Computer and Information Science  

As research momentum shifts toward the next evolution of generative AI (GenAI), Agentic AI, educational technologies are moving beyond reactive tools toward proactive, teammate-like ecosystems grounded in pedagogical principles. This transition raises a central challenge: how to design increasingly autonomous AI systems without diminishing learner agency or undermining teachers’ professional judgment. This workshop introduces the concept of HAI-Agency, envisioning how human and AI agency can be orchestrated in learning and teaching. Foregrounding proactive and reflective learning, we aim to advance a shared research agenda spanning design methodologies, computational modeling, evaluation frameworks, and the classroom integration of agentic AI systems. This workshop is of particular interest to the AIED community, especially researchers and practitioners across human-AI interaction, learning analytics, AI-supported learning design, and learning sciences. Through an interactive, hybrid, half-day format, we will explore how to shape a future in which agentic AI contributes to more equitable, transparent, and pedagogically sound educational innovations.

DOI： 10.1007/978-3-032-29794-5_5

Scopus

Publisher：Frontiers in Computer Science  

Introduction – Despite advances in virtual reality (VR) devices, leaders in VR-based group discussions lack inclusive facilitation support. Speaking-intention cues are latent information indicating readiness to speak, yet they are difficult to perceive in VR. Although existing works show the possibility of detecting speaking-intention cues using sensors and machine learning, a key question remains: how should such cues be delivered to align with leaders' perceived needs? Methods – We conducted a study of VR-based group discussion (N = 24) combining physiological sensing, behavioral coding of interaction dynamics, and post-hoc leader annotations and questionnaires. Results – Results show that leaders most often desired feedback during relaxed baseline states with short-term physiological fluctuations, indicating active cognitive regulation rather than high stress or complete stability. Leaders preferred feedback not only during observational phases but also during leader-dominant facilitation. Questionnaire results further reveal a strong preference for duration-based intention cues and generally non-anonymous feedback, as well as concerns about information overload and social pressure. Discussion – We finally discuss design implications for speaking-intention feedback in future VR systems supporting leadership and inclusive collaboration.

DOI： 10.3389/fcomp.2026.1794972

Web of Science

Scopus

Publisher：Proceedings of the ACM Symposium on Applied Computing  

University students often face financial challenges such as unstable incomes, rising living costs, and limited experience with money management, which can lead to stress, debt, and reduced academic engagement. Although digital budgeting tools are frequently recommended, many existing solutions fail to reflect students' diverse financial needs, motivations, and cultural backgrounds. To explore the design space for more inclusive student budgeting tools, we investigate how university students differ in their use of budgeting tools using survey data from 305 participants. Our analysis reveals clear behavioral differences shaped by saving goals, spending habits, and cultural context, and a clustering analysis identifies three distinct student profiles characterized by financial planning behaviors, income structures, and repayment practices. These findings highlight opportunities for designing budgeting tools that better support underserved and low-engagement student groups.

DOI： 10.1145/3748522.3779773

Scopus

Publisher：16th International Learning Analytics and Knowledge Conference Lak 2026  

As digital materials proliferate in higher education, e-book interaction logs provide a scalable lens on how students study. However, most existing research analyzes these logs in a one-dimensional manner, which limits the ability to capture students’ study approaches comprehensively. Moreover, the relationship between students’ study approaches, metacognition, and performance remains unclear. To address these challenges, we propose a three-dimensional framework that incorporates engagement, navigation pattern, and context, combining theory-driven and data-driven perspectives to define behavior-based features and group students with similar patterns. We apply this approach to data from a real-world class in which students used an e-book system to study course materials and complete comprehension quizzes. Our analysis identified three distinct groups of students with different study approaches, and revealed that engagement investment alone does not guarantee achievement. We further examined how these approaches relate to students’ metacognitive awareness and academic performance.

DOI： 10.1145/3785022.3785060

Scopus

Publishing type：Research paper (scientific journal)   Publisher：Journal of Intelligent Information Systems  

Recommender systems often face challenges related to information bias and the "filter bubble" effect, which limit content diversity and hinder user satisfaction. While much research has explored the balance between diversity and accuracy, traditional models often overlook the fact that different users have different levels of tolerance for diversity. In this work, we propose a novel approach PDRec-KG, which leverages knowledge graphs to enrich item representations with contextual information and integrate this with users’ historical interactions to estimate their personalized diversity tolerance. Building upon this estimation, we develop a framework that dynamically adjusts the trade-off between diversity and accuracy to align with each user’s preferences. Our method effectively tailors recommendations to individual diversity needs, offering a promising direction for enhancing both the quality and fairness of recommender systems. Comprehensive experiments on real-world datasets demonstrate the effectiveness of our approach. For instance, on the Last.FM dataset, PDRec-KG improves entity coverage by over 24% compared to the strongest baseline, while maintaining competitive accuracy, showcasing its ability to balance the trade-off between diversity and accuracy through personalized diverse recommendations.

DOI： 10.1007/s10844-025-01013-8

Web of Science

Scopus

researchmap

Authorship：Lead author,　Last author,　Corresponding author  

Publishing type：Research paper (scientific journal)  

DOI： 10.48550/arXiv.2603.22672

researchmap

Publishing type：Research paper (scientific journal)  

DOI： 10.48550/arXiv.2603.22674

researchmap

Publishing type：Research paper (scientific journal)  

DOI： 10.48550/arXiv.2603.22673

researchmap

Publishing type：Research paper (scientific journal)  

DOI： 10.48550/arXiv.2603.23830

researchmap

Publishing type：Research paper (scientific journal)   Publisher：ACM Transactions on Knowledge Discovery from Data  

Language learning applications usually estimate the learner’s language knowledge over time to provide personalized practice content for each learner at the optimal timing. However, accurately predicting language knowledge or linguistic skills is much more challenging than math or science knowledge, as many language tasks involve memorization and retrieval. Learners must memorize a large number of words and meanings, which are prone to be forgotten without practice. Although a few studies consider forgetting when modeling learners’ language knowledge, they tend to apply traditional models, consider only partial information about forgetting, and ignore linguistic features that may significantly influence learning and forgetting. This article focuses on modeling and predicting learners’ knowledge by considering their forgetting behavior and linguistic features in language learning. Specifically, we first explore the existence of forgetting behavior and cross-effects in real-world language learning datasets through empirical studies. Based on these, we propose a model for predicting the probability of recalling a word given a learner’s practice history. The model incorporates (1) three types of key information related to forgetting (time-gap, interaction, and word features), (2) question formats, and (3) similarities between words using the attention mechanism. Extensive experiments on two real-world datasets show that the proposed model improves performance compared to baselines. Moreover, the results indicate that combining multiple types of forgetting information and item format improves performance. In addition, we find that incorporating semantic and morphological features, such as word embeddings, to model similarities between words in a learner’s practice history and their effects on memory also improves the model. Our work indicates a potential future research direction for the knowledge tracing task in second language acquisition, which gives more instructive results for enhancing learning and teaching.

DOI： 10.1145/3778163

Web of Science

Scopus

researchmap

Publishing type：Part of collection (book)   Publisher：Springer Nature Switzerland  

DOI： 10.1007/978-3-032-05817-1_5

researchmap

Publishing type：Research paper (scientific journal)  

DOI： 10.1109/TCSS.2026.3716220

researchmap

Publisher：Communications in Computer and Information Science  

Despite the expansion of global connectivity, approximately 2.5 billion people in rural or developing regions remain offline, a digital divide that can significantly worsen educational inequality as learning increasingly relies on smart technologies. To bridge this gap, we propose Distributed Cooperative Learning Environments for Development (DCL4D), a framework that supports learners in environments without constant internet access. By integrating advanced techniques such as delay-tolerant networking, distributed data sharing, federated and progressive analytics, the framework can be used to provide intelligent support for different kinds of learning activities. These developments can help students in disconnected communities to benefit from personalized, data-driven educational tools, marking a critical first step toward making digital learning accessible to all, regardless of their infrastructure or location.

DOI： 10.1007/978-3-032-30833-7_6

Scopus

Publishing type：Research paper (scientific journal)   Publisher：Big Data and Cognitive Computing  

Massive Open Online Courses (MOOCs) have gained increasing popularity in recent years, highlighting the growing importance of effective course recommendation systems (CRS). However, the performance of existing CRS methods is often limited by data sparsity and suffers under cold-start scenarios. One promising solution is to leverage course-level conceptual information as side information to enhance recommendation performance. We propose a general framework for integrating LLM-generated concepts as side information into various classic recommendation algorithms. Our framework supports multiple integration strategies and is evaluated on two real-world MOOC datasets, with particular focus on the cold-start setting. The results show that incorporating LLM-generated concepts consistently improves recommendation quality across diverse models and datasets, demonstrating that automatically generated semantic information can serve as an effective, reusable, and scalable source of side knowledge for educational recommendations. This finding suggests that LLMs can function not merely as content generators but as practical data augmenters, offering a new direction for enhancing robustness and generalizability in course recommendation.

DOI： 10.3390/bdcc9120311

Web of Science

Scopus

researchmap

Authorship：Lead author,　Corresponding author   Language：English   Publishing type：Research paper (international conference proceedings)  

Authorship：Last author,　Corresponding author   Language：English   Publishing type：Research paper (international conference proceedings)  

Publishing type：Research paper (scientific journal)  

DOI： 10.48550/arXiv.2511.04144

researchmap

Publishing type：Research paper (scientific journal)   Publisher：Machine Learning and Knowledge Extraction  

Students must navigate large catalogs of courses and make appropriate enrollment decisions in many online learning environments. In this context, identifying key concepts and their relationships is essential for understanding course content and informing course recommendations. However, identifying and extracting concepts can be an extremely labor-intensive and time-consuming task when it has to be done manually. Traditional NLP-based methods to extract relevant concepts from courses heavily rely on resource-intensive preparation of detailed course materials, thereby failing to minimize labor. As recent advances in large language models (LLMs) offer a promising alternative for automating concept identification and relationship inference, we thoroughly investigate the potential of LLMs in automatically generating course concepts and their relations. Specifically, we systematically evaluate three LLM variants (GPT-3.5, GPT-4o-mini, and GPT-4o) across three distinct educational tasks, which are concept generation, concept extraction, and relation identification, using six systematically designed prompt configurations that range from minimal context (course title only) to rich context (course description, seed concepts, and subtitles). We systematically assess model performance through extensive automated experiments using standard metrics (Precision, Recall, F1, and Accuracy) and human evaluation by four domain experts, providing a comprehensive analysis of how prompt design and model choice influence the quality and reliability of the generated concepts and their interrelations. Our results show that GPT-3.5 achieves the highest scores on quantitative metrics, whereas GPT-4o and GPT-4o-mini often generate concepts that are more educationally meaningful despite lexical divergence from the ground truth. Nevertheless, LLM outputs still require expert revision, and performance is sensitive to prompt complexity. Overall, our experiments demonstrate the viability of LLMs as a tool for supporting educational content selection and delivery.

DOI： 10.3390/make7030103

Web of Science

Scopus

researchmap

Authorship：Corresponding author   Language：English   Publishing type：Research paper (international conference proceedings)  

Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

E-books generate extensive log data that sheds light on student behaviors. Among these, page jumps offer unique insights into reading strategies. However, prior research seldom connects these jumps to both the e-book's content and page functions and often presents only fragmented information. This study investigates how e-book content and page types relate to student page jump behaviors. We also propose a visualization framework that integrates e-book content and log data, enabling intuitive reading path visualizations and detailed analysis of student interactions. Our approach aims to offer educators actionable insights for refining instructional materials and providing more personalized feedback, ultimately enhancing the e-book learning experience.

DOI： 10.1109/icalt64023.2025.00038

Scopus

researchmap

Publishing type：Research paper (scientific journal)  

DOI： 10.48550/arXiv.2504.08856

researchmap

Publishing type：Research paper (scientific journal)  

DOI： 10.48550/arXiv.2504.08208

researchmap

Authorship：Lead author,　Last author,　Corresponding author  

researchmap

researchmap

Authorship：Last author,　Corresponding author  

researchmap

Publisher：Lecture Notes in Computer Science  

The spaced repetition technique is widely used in language learning applications to improve long-term memory retention in learners. However, most traditional algorithms for spaced repetition are simple functions with a few parameters and lack temporal information to model the forgetting process, and recent reinforcement learning methods utilize model-free algorithms where student memory retention is not estimated. Consequently, these models are inadequate at estimating a learner’s performance and unable to be used adaptively based on user feedback, which results in unsatisfactory review schedules for the learner. To address this issue, this research proposes a personalized language learning framework that utilizes deep learning to design a trainable, adaptive, and efficient spaced repetition scheduling method. Specifically, the framework includes a forgetting-aware knowledge tracing model to track students’ memory and a reinforcement learning based spaced repetition scheduling algorithm to achieve greater memorization efficiency. Experiments based on real-world datasets have shown performance improvement of the proposed approach over other baseline methods.

DOI： 10.1007/978-3-031-98459-4_19

Web of Science

Scopus

researchmap

Publisher：Lecture Notes in Computer Science  

The rapid advancement of AI has created transformative opportunities in education, reshaping how students learn. While much of the existing research has focused on AI’s impact on student behavior and performance, its effects on emotional experiences and learning engagement remain underexplored. This study aims to address this gap using questionnaires and video analysis to capture students’ real-time emotions and engagement levels, investigating how AI influences these two dimensions in programming education. Our preliminary results reveal interesting findings and lay the foundation for further research.

DOI： 10.1007/978-3-031-98462-4_30

Web of Science

Scopus

researchmap

Publisher：Lecture Notes in Computer Science  

In learning analytics, early risk prediction plays a critical role in identifying students who are at risk of academic failure, enabling timely interventions to improve student outcomes. While effective in predictive accuracy, traditional machine learning models often require the training of multiple models for different time frames and suffer from a lack of explainability, limiting their practical application in educational settings. In this research, we propose ensemble dendritic neuron models (EDNMs), a novel approach for early risk prediction that addresses the challenges of existing models. EDNMs take advantage of a neural pruning mechanism inspired by biological neurons, allowing visual feature selection and explainability. The proposed EDNMs inherent visual feature selection improves transparency, making it easier for educators to interpret which factors contribute to the risk of a student. The performance of EDNMs is evaluated against RNN-based methods, demonstrating superior efficiency in prediction tasks and improved explainability. Crucially, the performance of early predictions over different timeframes is also provided by dynamically adjusting the dendritic state. Unlike conventional models that require multiple versions to accommodate predictions at different timeframes, EDNMs can adjust to varying weeks of early prediction with a single model, significantly reducing computational overhead. This study contributes to the growing field of explainable AI in education by offering a practical solution that enhances both the efficiency and transparency of early risk prediction systems.

DOI： 10.1007/978-3-031-98417-4_13

Web of Science

Scopus

researchmap

Publisher：Ceur Workshop Proceedings  

Concept maps have been widely used in education to organize and represent information hierarchically. However, traditional methods for constructing concept maps often depend on human experts, which can be costly and time-consuming. The emergence of large language models (LLMs), such as GPT-4, has transformed concept construction and reasoning tasks by offering automated and scalable solutions. This paper introduces a novel framework for generating concept maps of e-books with three key components: section segmentation, key concept extraction, and relationship identification. Additionally, the paper highlights challenges and future opportunities to enhance LLM-driven concept map generation for educational applications.

Scopus

Publisher：Proceedings 25th IEEE International Conference on Advanced Learning Technologies Icalt 2025  

Traditional knowledge graph (KG) approach often rely on static textbook content and overlook the dynamic, collaborative interactions in collaborative problem-solving (CPS). This study introduced a three-step integrated KG approach designed to support CPS in STEM education and examined the effective KG features that influence CPS learning outcomes. KGs were generated by combining learning materials and student dialogue data. Two types of features, graph structural features and knowledge activation features, were identified to classify knowledge nodes and analyze how students activated knowledge during CPS. Clustering analysis revealed three types of knowledge nodes: Peripheral Nodes, Core Nodes, and Degree Hubs. Furthermore, key features such as depth, branch, and activated paths showed positive correlations with group discussion performance and CPS skills but had limited influence on test scores. These findings highlight the potential of integrated KGs to support both individual and group learning in STEM education.

DOI： 10.1109/ICALT64023.2025.00036

Scopus

Publishing type：Research paper (scientific journal)   Publisher：Applied Sciences Switzerland  

This study explores the factors influencing turn-taking coordination in virtual reality (VR) environments, with a focus on identifying key interaction dynamics that affect the ease of gaining speaking opportunities. By analyzing VR interaction data through logistic regression and clustering, we identify significant variables impacting turn-taking success and categorize typical interaction states that present unique coordination challenges. The findings reveal that features related to interaction proactivity, individual status, and communication quality significantly impact turn-taking outcomes. Furthermore, clustering analysis identifies five primary interaction contexts: high competition, intense interaction, prolonged single turn, high-status role, and low activity, each with unique turn-taking coordination requirements. This work provides insights into enhancing turn-taking support systems in VR, emphasizing contextually adaptive feedback to reduce speaking overlap and turn-taking failures, thereby improving overall interaction flow in immersive environments.

DOI： 10.3390/app142412071

Web of Science

Scopus

researchmap

Language：Others  

This study investigates the effects of generative AI on the knowledge comprehension of university students, focusing on the use of ChatGPT strategies. Data from 81 junior students who used the ChatGPT worksheet were collected and analyzed. Path analysis revealed complex interactions between ChatGPT strategy use, e-book reading behaviors, and students' prior perceived understanding of concepts. Students' prior perceived understanding and reading behaviors indirectly affected their final scores, mediated by the ChatGPT strategy use. The mediation effects indicated that reading behaviors significantly influenced final scores through ChatGPT strategies, indicating the importance of the interaction with learning materials. Further regression analysis identified the specific ChatGPT strategy related to verifying and comparing information sources as significantly influenced by reading behaviors and directly affecting students' final scores. The findings provide implications for practical strategic guidance for integrating ChatGPT in education.

Language：Others  

The integration of ChatGPT as a supportive tool in education, notably in programming courses, addresses the unique challenges of programming education by providing assistance with debugging, code generation, and explanations. Despite existing research validating ChatGPT’s effectiveness, its application in university-level programming education and a detailed understanding of student interactions and perspectives remain limited. This paper explores ChatGPT’s impact on learning in a Python programming course tailored for first-year students over eight weeks. By analyzing responses from surveys, open-ended questions, and student-ChatGPT dialog data, we aim to provide a comprehensive view of ChatGPT’s utility and identify both its advantages and limitations as perceived by students. Our study uncovers a generally positive reception toward ChatGPT and offers insights into its role in enhancing the programming education experience. These findings contribute to the broader discourse on AI’s potential in education, suggesting paths for future research and application.

Publishing type：Research paper (international conference proceedings)   Publisher：International Educational Data Mining Society  

researchmap

Other Link： https://dblp.uni-trier.de/rec/conf/edm/2024

Publishing type：Research paper (international conference proceedings)   Publisher：Springer  

This study investigates the effects of generative AI on the knowledge comprehension of university students, focusing on the use of ChatGPT strategies. Data from 81 junior students who used the ChatGPT worksheet were collected and analyzed. Path analysis revealed complex interactions between ChatGPT strategy use, e-book reading behaviors, and students’ prior perceived understanding of concepts. Students’ prior perceived understanding and reading behaviors indirectly affected their final scores, mediated by the ChatGPT strategy use. The mediation effects indicated that reading behaviors significantly influenced final scores through ChatGPT strategies, indicating the importance of the interaction with learning materials. Further regression analysis identified the specific ChatGPT strategy related to verifying and comparing information sources as significantly influenced by reading behaviors and directly affecting students’ final scores. The findings provide implications for practical strategic guidance for integrating ChatGPT in education.

DOI： 10.1007/978-3-031-64315-6_12

Scopus

researchmap

Other Link： https://dblp.uni-trier.de/db/conf/aied/aied2024c1.html#ChenLMTOS24

Language：Others  

Language proficiency diagnosis is essential to extract fine-grained information about the linguistic knowledge states and skill mastery levels of test takers based on their performance on language tests. Different from comprehensive standardized tests, many language learning apps often revolve around word-level questions. Therefore, knowledge concepts and linguistic skills are hard to define, and diagnosis must be well-designed. Traditional approaches are widely applied for modeling knowledge in science or mathematics, where skills or knowledge concepts are easy to associate with each item. However, only a few works focus on defining knowledge concepts and skills using linguistic characteristics for language knowledge proficiency diagnosis. In addressing this, we propose a framework for language proficiency diagnosis based on neural networks. Specifically, we propose a series of methods based on our framework that uses different linguistic features to define skills and knowledge concepts in the context of the language learning task. Experimental results on a real-world second-language learning dataset demonstrate the effectiveness and interpretability of our framework. We also provide empirical evidence with comprehensive experiments and analysis to prove that our knowledge concept and skill definitions are reasonable and critical to the performance of our model.

Language：English  

Authorship：Lead author,　Last author,　Corresponding author   Publishing type：Research paper (international conference proceedings)   Publisher：Springer  

An emerging challenge in course recommendation systems is the need to explain clearly to students the rationale behind specific course recommendations. Consequently, recent research has transitioned from focusing primarily on the accuracy of these systems to prioritizing user-centric qualities, such as transparency and justification. This shift has led to an increased emphasis on methods that provide clear, understandable explanations for their recommendations. In response to this trend, our paper introduces an explainable recommendation framework. Utilizing this framework, we analyze existing course recommendation systems and explore the emerging research challenges and future prospects for explainable course recommendation systems.

DOI： 10.1007/978-3-031-60012-8_17

Web of Science

Scopus

researchmap

Other Link： https://dblp.uni-trier.de/db/conf/hci/hci2024-36.html#MaYR24

researchmap

researchmap

Authorship：Lead author,　Last author,　Corresponding author  

researchmap

Language：Japanese   Publisher：Special Interest Group on Web Intelligence and Interaction  

<p>Affective states play a crucial role in music, as music can influence both current emotions and long-term moods. We propose a novel <i>emotion-enhancing</i> hybrid music recommendation (EEMR) mechanism that finds the suitable criteria in selecting the best music playlist for improving user’s emotion by combining two recommendation techniques, i.e., content-based filtering and context-aware approach. This mechanism generates playlists that align with the user’s preferences and current emotion, while also supporting gradual improvement of the user’s emotion over time.</p>

DOI： 10.57413/wii.20.0_79

CiNii Research

Web of Science

Publisher：Proceedings of the International Conference on Educational Data Mining  

Course recommender systems can assist students in identifying suitable or appealing courses by leveraging user interaction data, which shows previous engagements between users and courses. However, a prevalent issue with existing course recommender systems is their tendency to prioritize accuracy over explainability. The ”black-box” nature of these complex models presents a challenge: accurately characterizing and modeling users’ preferences while also providing explicit, comprehensive, and explainable user profiles. To address this limitation, we propose a novel Knowledge Entity-Aware Model for course recommendation called KEAM, which supports explicit user profile generation based on detailed information from a knowledge graph to enhance students’ comprehension of the rationales behind the recommendations. Specifically, we exploit the information encoded in a knowledge graph to build connections between units using a neural network by replacing the hidden units. Next, the model is trained to capture students’ preferences and create user profiles for explainable recommendations. Comprehensive experiments have been conducted on two real-world online datasets to evaluate the proposed model’s effectiveness and explainability.

DOI： 10.5281/zenodo.12729910

Scopus

CiNii Research

CiNii Research

CiNii Research

Publisher：Proceedings of the 21st International Conference on Cognition and Exploratory Learning in the Digital Age Celda 2024  

This paper proposes a three-step approach to develop knowledge graphs that integrate textbook-based target knowledge graph with student dialogue-based knowledge graphs. The study was conducted in seventh-grade STEM classes, following a collaborative problem solving process. First, the proposed approach generates a comprehensive target knowledge graph from learning material contents, establishing a reference framework that represents the target knowledge structure of the course. Second, customized knowledge graphs were generated by analyzing the scientific concepts and knowledge based on the discussion dialogues, showing students' activated knowledge structures. Finally, the dialogue-based knowledge graphs were integrated into textbook-based target knowledge, to identify the activated and non-activated knowledge nodes and connections, as well as the related activated knowledge nodes and connections from other previous lectures or experiences. This three-step approach visualizes students' knowledge activation, and the learning gaps remain. This paper presented three examples of integrated knowledge graphs based on the different group formations. The findings of three different groups were discussed, and some educational implications were provided.

Scopus

Publisher：Proceedings of the 21st International Conference on Cognition and Exploratory Learning in the Digital Age Celda 2024  

As education increasingly shifts towards a technology-driven model, artificial intelligence systems like ChatGPT are gaining recognition for their potential to enhance educational support. In university education and MOOC environments, students often select courses that align with their specific needs. During this process, access to information about the knowledge concepts covered in a course can help students make more informed decisions. However, manually constructing this knowledge concept information is a labor-intensive and time-consuming task. In this paper, we explore the capability of ChatGPT in generating relevant knowledge concepts from course syllabi and evaluate the accuracy and consistency of these AI-generated concepts against course content using four assessment techniques at both the concept level and course level. We investigate the feasibility of using ChatGPT-generated concepts as a direct educational resource, as well as their potential integration into broader educational technologies, such as interpretable course recommendation systems.

Scopus

Publisher：Proceedings of the 21st International Conference on Cognition and Exploratory Learning in the Digital Age Celda 2024  

Generative artificial intelligence (AI) tools like ChatGPT are becoming increasingly common in educational settings, especially in programming education. However, the impact of these tools on the learning process, student performance, and best practices for their integration remains underexplored. This study examines student experiences and interactions using ChatGPT in a beginner-level Python programming course through a combination of questionnaire responses and student-ChatGPT dialogue data analysis. The findings reveal a generally positive student reception toward ChatGPT, emphasizing its role in enhancing the programming education experience. Additionally, by clustering and analyzing the types of prompts students use, we identify four distinct patterns of ChatGPT usage and compare the performance outcomes associated with each pattern. This empirical research provides a deeper understanding of AI-enhanced programming education, offering valuable insights and suggesting pathways for future research and practical applications.

Scopus

Publisher：Proceedings of the International Conference on Educational Data Mining  

Vocabulary proficiency diagnosis plays an important role in the field of language learning, which aims to identify the level of vocabulary knowledge of a learner through his or her learning process periodically, and can be used to provide personalized materials and feedback in language-learning applications. Traditional approaches are widely applied for modeling knowledge in science or mathematics, where skills or knowledge concepts are well-defined and easy to associate with each item. However, only a handful of works focus on defining knowledge concepts and skills using linguistic characteristics for language knowledge proficiency diagnosis. In addressing this, we propose a framework for vocabulary proficiency diagnosis based on neural networks. Specifically, we propose a series of methods based on our framework that uses different linguistic features to define skills and knowledge concepts in the context of the language learning task. Experimental results on a real-world second-language learning dataset demonstrate the effectiveness and interpretability of our framework. We also provide empirical evidence with ablation testing to prove that our knowledge concept and skill definitions are reasonable and critical to the performance of our model.

DOI： 10.5281/zenodo.8115675

Scopus

Authorship：Lead author,　Last author,　Corresponding author   Language：English   Publishing type：Research paper (international conference proceedings)  

researchmap

Language：Others   Publishing type：Research paper (other academic)  

The COVID-19 pandemic has resulted in school closures all across the world, and lots of students have shifted from conventional classrooms to online learning. With the help of ICT technologies nowadays, learning online can be more effective in a number of ways. However, most of the online learning environments without instructors' attention may result in different learning patterns compared to the traditional face-to-face classroom. In this paper, we aimed at detecting the slide reading behaviors of the students by analyzing operational event logs from a digital textbook reader for a lecture offered in our university. We compared reading patterns between traditional face-to-face lectures and hybrid online lectures, our results show that online lectures lead to more off-task behaviors. Our analysis provides a rich understanding of e-book reading and informs design implications for online learning during the pandemic. The findings can also be used to improve the instruction designs and learning strategies.

Language：Others   Publishing type：Research paper (other academic)  

The abundance of courses available in university and the highly personalized curriculum is often overwhelming for students who must select courses relevant to their academic interests. A large body of research in course recommendation systems focuses on optimizing prediction and improving accuracy. However, those systems usually afford little or no user interaction, and little is known about the influence of user-perceived aspects for course recommendations, such as transparency, controllability, and user satisfaction. In this paper, we argue that involving students in the course recommendation process is important, and we present an interactive course recommendation system that provides explanations and allows students to explore courses in a personalized way. A within-subject user study was conducted to evaluate our system and the results show a significant improvement in many user-centric metrics.

Language：Others  

Language：Others  

Language：Others   Publishing type：Research paper (other academic)  

Course recommendation system is a useful tool that not only helps the students who have no sufficient experience to decide what they should study, but also leverages their full performance if they could study what they like or are interested in. Different from MOOCs, the selection and recommendation for hybrid learning environments such as university are relatively difficult. Students who enrolled in the same course may have completely different purposes and different interest. Employing the enrollment record data from Kyushu University, we conduct a systematic investigation on the course-taking pattern for recommendation. We then discuss the challenges to recommend suitable courses in university and propose a preliminary approach to address the challenges by designing a course recommendation mechanism based on association rule of previous course-taking pattern together with student interest.

researchmap

researchmap

researchmap

Type：Peer review 

Number of peer-reviewed articles in foreign language journals：6

Proceedings of International Conference Number of peer-reviewed papers：4

Type：Peer review 

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap