Pub Date : 2025-10-17DOI: 10.1109/TLT.2025.3622946
Ye Zhang;Yupeng Zhou;Jun Wu;Wenbo Zhou;Xiaowei Zhao;Weiwei Wang;Minghao Yin
Personalized exercise recommendation is essential for enhancing learning efficiency by adapting educational content to individual student needs. However, current approaches often fail to simultaneously minimize the number of exercises while ensuring that students achieve specific learning objectives, and they overlook the nuanced knowledge structures revealed during problem solving. To address these limitations, this article introduces a novel framework for dynamic requirement-driven exercise recommendation (DRER), which combines confusion-aware knowledge tracing with nonlinear combinatorial optimization. DRER consists of two key stages. First, in the knowledge tracing stage, a hybrid knowledge concept matrix is constructed to model both the inherent relationships among knowledge concepts and the latent structures revealed through student interactions. Second, in the exercise recommendation stage, the problem is formulated as a minimum nonlinear weighted set cover problem, aiming to identify the smallest set of exercises that enables students to reach a predefined proficiency threshold (e.g., 0.6). To solve this efficiently, a heuristic-based local search algorithm is proposed. Extensive experiments on real-world and academic datasets validate the effectiveness of the framework, demonstrating its ability to significantly reduce the number of recommended exercises while ensuring high accuracy in achieving learning objectives. This work represents a significant integration of data mining and combinatorial optimization, offering a scalable and practical solution for personalized education.
{"title":"Dynamic Requirement-Driven Exercise Recommendation via Confusion-Aware Knowledge Tracing and Nonlinear Combinatorial Optimization","authors":"Ye Zhang;Yupeng Zhou;Jun Wu;Wenbo Zhou;Xiaowei Zhao;Weiwei Wang;Minghao Yin","doi":"10.1109/TLT.2025.3622946","DOIUrl":"https://doi.org/10.1109/TLT.2025.3622946","url":null,"abstract":"Personalized exercise recommendation is essential for enhancing learning efficiency by adapting educational content to individual student needs. However, current approaches often fail to simultaneously minimize the number of exercises while ensuring that students achieve specific learning objectives, and they overlook the nuanced knowledge structures revealed during problem solving. To address these limitations, this article introduces a novel framework for dynamic requirement-driven exercise recommendation (<sc>DRER</small>), which combines confusion-aware knowledge tracing with nonlinear combinatorial optimization. <sc>DRER</small> consists of two key stages. First, in the knowledge tracing stage, a hybrid knowledge concept matrix is constructed to model both the inherent relationships among knowledge concepts and the latent structures revealed through student interactions. Second, in the exercise recommendation stage, the problem is formulated as a minimum nonlinear weighted set cover problem, aiming to identify the smallest set of exercises that enables students to reach a predefined proficiency threshold (e.g., 0.6). To solve this efficiently, a heuristic-based local search algorithm is proposed. Extensive experiments on real-world and academic datasets validate the effectiveness of the framework, demonstrating its ability to significantly reduce the number of recommended exercises while ensuring high accuracy in achieving learning objectives. This work represents a significant integration of data mining and combinatorial optimization, offering a scalable and practical solution for personalized education.","PeriodicalId":49191,"journal":{"name":"IEEE Transactions on Learning Technologies","volume":"18 ","pages":"989-1002"},"PeriodicalIF":4.9,"publicationDate":"2025-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145612120","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-15DOI: 10.1109/TLT.2025.3622043
Qian Fu;Yaning Zhao;Zixi Jia;Yafeng Zheng
Large language models (LLMs) show great potential in programming learning. However, existing studies mainly focus on technical implementations and lack a systematic analysis of the application of LLMs in programming learning from an educational perspective. This study conducts a systematic literature review and bibliometric analysis based on 75 high-quality papers, using a 6-D framework (roles, technology, learners, environment, effectiveness, and challenges) to examine the current state and agenda of LLM applications. The results indicate that the application of LLMs has evolved from model validation in 2022 to teaching applications in 2023 and is expected to be deeply integrated into the education system by 2024–2025, reflecting a shift from tools to teaching agents. In programming learning, LLMs primarily take on roles in resource generation, task solving, and feedback provision. In terms of technology usage, OpenAI’s series of models dominate, with Python being the main programming language environment, and research subjects focusing on beginner programmers and university students. Empirical studies show that LLMs can effectively enhance learners’ cognitive outcomes and noncognitive performance, but they can also lead to overreliance on tools, academic integrity risks, and ethical challenges. Future research should establish an education theory-driven design framework for LLMs, conduct studies on generative artificial intelligence literacy and ethical norms, and provide theoretical and practical guidance for programming learning.
{"title":"Large Language Models (LLMs) in Programming Learning: The Current Research State and Agenda","authors":"Qian Fu;Yaning Zhao;Zixi Jia;Yafeng Zheng","doi":"10.1109/TLT.2025.3622043","DOIUrl":"https://doi.org/10.1109/TLT.2025.3622043","url":null,"abstract":"Large language models (LLMs) show great potential in programming learning. However, existing studies mainly focus on technical implementations and lack a systematic analysis of the application of LLMs in programming learning from an educational perspective. This study conducts a systematic literature review and bibliometric analysis based on 75 high-quality papers, using a 6-D framework (roles, technology, learners, environment, effectiveness, and challenges) to examine the current state and agenda of LLM applications. The results indicate that the application of LLMs has evolved from model validation in 2022 to teaching applications in 2023 and is expected to be deeply integrated into the education system by 2024–2025, reflecting a shift from tools to teaching agents. In programming learning, LLMs primarily take on roles in resource generation, task solving, and feedback provision. In terms of technology usage, OpenAI’s series of models dominate, with Python being the main programming language environment, and research subjects focusing on beginner programmers and university students. Empirical studies show that LLMs can effectively enhance learners’ cognitive outcomes and noncognitive performance, but they can also lead to overreliance on tools, academic integrity risks, and ethical challenges. Future research should establish an education theory-driven design framework for LLMs, conduct studies on generative artificial intelligence literacy and ethical norms, and provide theoretical and practical guidance for programming learning.","PeriodicalId":49191,"journal":{"name":"IEEE Transactions on Learning Technologies","volume":"18 ","pages":"942-961"},"PeriodicalIF":4.9,"publicationDate":"2025-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145510178","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Educational process mining has demonstrated significant promise in predicting learners’ academic performance. However, for interdisciplinary courses, traditional centralized modeling methods might blur the inherent differences in ability profiles and knowledge frameworks among learners from various disciplines. Therefore, this article proposes a framework of enhanced capability and privacy for interdisciplinary course performance prediction in higher education (ARISE), which is based on personalized federal learning (PFL) and deep learning (DL). We design a lightweight DL network, divided into nonpersonalized and personalized layers that, respectively, preserve common information from learners’ course studies and their discipline-specific capabilities. To obtain optimal predictive factors from abundant prerequisite courses, we propose a reinforcement-learning-based feature selection method, utilizing its capability to adapt to data variations and reward mechanism to choose influential prerequisite courses. In addition, the distributed collaborative training and the strategy of not uploading personalized layer parameters in PFL can achieve dual privacy protection for learners, thereby improving the security of ARISE. We conduct extensive experiments on three datasets, achieving state-of-the-art results. The average accuracy of PFL across clients for each dataset is 70.12%, 89.45%, and 69.14%.
{"title":"ARISE: Enhanced Capability and Privacy for Interdisciplinary Course Performance Prediction in Higher Education","authors":"Fang Liu;Xinyue Chen;Qin Dai;Chunlong Fan;Jun Shen;Liang Zhao","doi":"10.1109/TLT.2025.3620870","DOIUrl":"https://doi.org/10.1109/TLT.2025.3620870","url":null,"abstract":"Educational process mining has demonstrated significant promise in predicting learners’ academic performance. However, for interdisciplinary courses, traditional centralized modeling methods might blur the inherent differences in ability profiles and knowledge frameworks among learners from various disciplines. Therefore, this article proposes a framework of enhanced c<underline>a</u>pability and p<underline>r</u>ivacy for <underline>i</u>nterdisciplinary cour<underline>s</u>e performance pr<underline>e</u>diction in higher education (ARISE), which is based on personalized federal learning (PFL) and deep learning (DL). We design a lightweight DL network, divided into nonpersonalized and personalized layers that, respectively, preserve common information from learners’ course studies and their discipline-specific capabilities. To obtain optimal predictive factors from abundant prerequisite courses, we propose a reinforcement-learning-based feature selection method, utilizing its capability to adapt to data variations and reward mechanism to choose influential prerequisite courses. In addition, the distributed collaborative training and the strategy of not uploading personalized layer parameters in PFL can achieve dual privacy protection for learners, thereby improving the security of ARISE. We conduct extensive experiments on three datasets, achieving state-of-the-art results. The average accuracy of PFL across clients for each dataset is 70.12%, 89.45%, and 69.14%.","PeriodicalId":49191,"journal":{"name":"IEEE Transactions on Learning Technologies","volume":"18 ","pages":"928-941"},"PeriodicalIF":4.9,"publicationDate":"2025-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145510179","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-06DOI: 10.1109/TLT.2025.3616515
Qing Li;Xin Yuan;Jianwen Sun;Xinrui Li;Sijing Chen;Xiaoxuan Shen;Sannyuya Liu
Deep-learning-based knowledge tracing (DLKT) models have achieved high predictive accuracy, but their opaque “black box” nature limits practical value: educators cannot trace why predictions are made and learners cannot obtain transparent feedback. Existing explainability techniques, which aim to explain why a model makes a particular prediction and provide the evidence for it, predominantly rely on correlational analyses, often yielding unfaithful or suboptimal explanations. To address this, we propose a post hoc reinforcement-learning-based causal deep knowledge tracing explainer (RCKTE). RCKTE operates through a structured workflow: First, it formulates the explanation task as a globally optimal subsequence screening problem, aiming to identify the most causally influential historical interactions from a student’s learning sequence. Then, a reinforcement learning agent, guided by a causal attribution reward and a dual-optimizer scheme, iteratively constructs the optimal subsequences by assessing the causal impact of each interaction. This process results in more faithful and concise explainable subsequences than those produced by correlation-based methods, achieving this within about 1 s to support real-time use. Finally, these explainable subsequences directly support actionable educational applications, including identifying a learner’s weak knowledge for targeted review, constructing personalized knowledge structure graphs for intervention tracking, and deriving group-level knowledge structures to guide curriculum design. Extensive experiments across multiple DLKT models and datasets confirm that RCKTE consistently outperforms existing post hoc methods in both the faithfulness and readability of explanations. By integrating causal attribution with reinforcement learning, RCKTE provides accurate, efficient, and educationally meaningful explanations that enhance the usability of DLKT in real learning environments.
{"title":"RCKTE: Toward Global Optimal Causal Explanations for Deep Knowledge Tracing","authors":"Qing Li;Xin Yuan;Jianwen Sun;Xinrui Li;Sijing Chen;Xiaoxuan Shen;Sannyuya Liu","doi":"10.1109/TLT.2025.3616515","DOIUrl":"https://doi.org/10.1109/TLT.2025.3616515","url":null,"abstract":"Deep-learning-based knowledge tracing (DLKT) models have achieved high predictive accuracy, but their opaque “black box” nature limits practical value: educators cannot trace why predictions are made and learners cannot obtain transparent feedback. Existing explainability techniques, which aim to explain why a model makes a particular prediction and provide the evidence for it, predominantly rely on correlational analyses, often yielding unfaithful or suboptimal explanations. To address this, we propose a post hoc reinforcement-learning-based causal deep knowledge tracing explainer (RCKTE). RCKTE operates through a structured workflow: First, it formulates the explanation task as a globally optimal subsequence screening problem, aiming to identify the most causally influential historical interactions from a student’s learning sequence. Then, a reinforcement learning agent, guided by a causal attribution reward and a dual-optimizer scheme, iteratively constructs the optimal subsequences by assessing the causal impact of each interaction. This process results in more faithful and concise explainable subsequences than those produced by correlation-based methods, achieving this within about 1 s to support real-time use. Finally, these explainable subsequences directly support actionable educational applications, including identifying a learner’s weak knowledge for targeted review, constructing personalized knowledge structure graphs for intervention tracking, and deriving group-level knowledge structures to guide curriculum design. Extensive experiments across multiple DLKT models and datasets confirm that RCKTE consistently outperforms existing post hoc methods in both the faithfulness and readability of explanations. By integrating causal attribution with reinforcement learning, RCKTE provides accurate, efficient, and educationally meaningful explanations that enhance the usability of DLKT in real learning environments.","PeriodicalId":49191,"journal":{"name":"IEEE Transactions on Learning Technologies","volume":"18 ","pages":"914-927"},"PeriodicalIF":4.9,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145405309","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-03DOI: 10.1109/TLT.2025.3617909
Juan Yang;Minjuan Wang;Xu Du;Rina Na
Modern education aims at providing students with more personalized learning services and more engaging learning experiences. One promising approach is to develop educational agents to facilitate high-quality completion of various educational tasks. In recent years, the advent of large language models (LLMs) has breathed new life into educational agents and pushed them into a new stage of intelligence. This survey tries to conduct a comprehensive and thorough investigation of LLM-based agents in education. First, the developments of educational agents are presented as background information. Subsequently, we propose a unified architecture for LLM-based educational agents, including perception, profiling, memory, reasoning, and action modules, and summarize two primary methods (i.e., fine-tuning and prompt engineering) for equipping them with abilities. Next, we categorize the potential applications of LLM-based educational agents across the “teaching–learning–assessment–research” chain, and discover that LLM-based educational agent can play significant roles in various educational tasks. Furthermore, we reveal that when assessing the effectiveness of LLM-based educational agents, subjective evaluation remains dominant, supplemented by objective evaluation. Finally, the open issues and future research directions in this field are discussed from multiple perspectives. We hope that this survey can provide valuable insights and inspirations for researchers and practitioners to enhance the further development of educational agents in the future.
{"title":"A Comprehensive Survey on Large-Language-Model-Based Agents for Education","authors":"Juan Yang;Minjuan Wang;Xu Du;Rina Na","doi":"10.1109/TLT.2025.3617909","DOIUrl":"https://doi.org/10.1109/TLT.2025.3617909","url":null,"abstract":"Modern education aims at providing students with more personalized learning services and more engaging learning experiences. One promising approach is to develop educational agents to facilitate high-quality completion of various educational tasks. In recent years, the advent of large language models (LLMs) has breathed new life into educational agents and pushed them into a new stage of intelligence. This survey tries to conduct a comprehensive and thorough investigation of LLM-based agents in education. First, the developments of educational agents are presented as background information. Subsequently, we propose a unified architecture for LLM-based educational agents, including perception, profiling, memory, reasoning, and action modules, and summarize two primary methods (i.e., fine-tuning and prompt engineering) for equipping them with abilities. Next, we categorize the potential applications of LLM-based educational agents across the “teaching–learning–assessment–research” chain, and discover that LLM-based educational agent can play significant roles in various educational tasks. Furthermore, we reveal that when assessing the effectiveness of LLM-based educational agents, subjective evaluation remains dominant, supplemented by objective evaluation. Finally, the open issues and future research directions in this field are discussed from multiple perspectives. We hope that this survey can provide valuable insights and inspirations for researchers and practitioners to enhance the further development of educational agents in the future.","PeriodicalId":49191,"journal":{"name":"IEEE Transactions on Learning Technologies","volume":"18 ","pages":"898-913"},"PeriodicalIF":4.9,"publicationDate":"2025-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145405250","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-16DOI: 10.1109/TLT.2025.3610636
Mingjing Huang;Ngai Cheong;Zhuofan Zhang;Jiaqi Liu
The rapid digitalization of higher education has positioned the Cognitive Evolution Engine (CEE)—defined as an adaptive artificial intelligence (AI) system that dynamically models learner cognition and evolves through iterative optimization—as an emerging technology for personalized learning in China. This systematic review examines CEE applications through comprehensive analysis of 14 953 documents from IEEE Xplore, Scopus, Web of Science, and CNKI (2014–2024), with 94 studies meeting rigorous criteria for detailed content analysis. Our analysis reveals both significant progress and implementation challenges. CEE research has grown exponentially since 2019, particularly following China’s 2022 Digital Education initiatives. However, we identified notable gaps between theoretical concepts and practical implementations, with most current systems utilizing established AI technologies rather than fully realized cognitive evolution mechanisms. Geographic distribution analysis indicates that 58% of high-quality research originates from eastern institutions, highlighting regional disparities in research capacity. In addition, system transparency emerged as a key concern, with a majority of empirical studies acknowledging challenges in algorithmic interpretability. Based on systematic synthesis, we propose a 3-D framework integrating technological infrastructure, pedagogical principles, and implementation strategies adapted to Chinese educational contexts. This framework provides guidance for advancing from current AI applications toward authentic CEE systems. Our research contributes the first comprehensive analysis of CEE in Chinese higher education, offering evidence-based insights for enhancing personalized learning while addressing identified implementation challenges.
高等教育的快速数字化将认知进化引擎（CEE）定位为中国个性化学习的新兴技术。认知进化引擎被定义为一种自适应人工智能（AI）系统，可以动态模拟学习者的认知，并通过迭代优化进行进化。本系统综述通过对来自IEEE Xplore、Scopus、Web of Science和CNKI（2014-2024）的14953篇文献的综合分析，对CEE应用进行了研究，其中94篇研究符合详细内容分析的严格标准。我们的分析揭示了重大进展和实施挑战。自2019年以来，中东欧研究呈指数级增长，特别是在中国2022年数字教育倡议之后。然而，我们发现了理论概念和实际实现之间的显著差距，大多数当前系统利用已建立的人工智能技术，而不是完全实现的认知进化机制。地理分布分析表明，58%的高质量研究来自东部机构，突出了研究能力的区域差异。此外，系统透明度成为一个关键问题，大多数实证研究承认算法可解释性方面的挑战。在系统综合的基础上，我们提出了一个整合技术基础设施、教学原则和实施策略的三维框架，以适应中国的教育环境。该框架为从当前的人工智能应用向真正的CEE系统发展提供了指导。我们的研究首次对中国高等教育中的CEE进行了全面分析，为加强个性化学习提供了基于证据的见解，同时解决了已确定的实施挑战。
{"title":"Exploring the Development of Personalizing Learning in Chinese Higher Education: A Systematic Review of Cognitive Evolution Engine by AI","authors":"Mingjing Huang;Ngai Cheong;Zhuofan Zhang;Jiaqi Liu","doi":"10.1109/TLT.2025.3610636","DOIUrl":"https://doi.org/10.1109/TLT.2025.3610636","url":null,"abstract":"The rapid digitalization of higher education has positioned the Cognitive Evolution Engine (CEE)—defined as an adaptive artificial intelligence (AI) system that dynamically models learner cognition and evolves through iterative optimization—as an emerging technology for personalized learning in China. This systematic review examines CEE applications through comprehensive analysis of 14 953 documents from IEEE Xplore, Scopus, Web of Science, and CNKI (2014–2024), with 94 studies meeting rigorous criteria for detailed content analysis. Our analysis reveals both significant progress and implementation challenges. CEE research has grown exponentially since 2019, particularly following China’s 2022 Digital Education initiatives. However, we identified notable gaps between theoretical concepts and practical implementations, with most current systems utilizing established AI technologies rather than fully realized cognitive evolution mechanisms. Geographic distribution analysis indicates that 58% of high-quality research originates from eastern institutions, highlighting regional disparities in research capacity. In addition, system transparency emerged as a key concern, with a majority of empirical studies acknowledging challenges in algorithmic interpretability. Based on systematic synthesis, we propose a 3-D framework integrating technological infrastructure, pedagogical principles, and implementation strategies adapted to Chinese educational contexts. This framework provides guidance for advancing from current AI applications toward authentic CEE systems. Our research contributes the first comprehensive analysis of CEE in Chinese higher education, offering evidence-based insights for enhancing personalized learning while addressing identified implementation challenges.","PeriodicalId":49191,"journal":{"name":"IEEE Transactions on Learning Technologies","volume":"18 ","pages":"877-897"},"PeriodicalIF":4.9,"publicationDate":"2025-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145352233","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-05DOI: 10.1109/TLT.2025.3606757
Jiutong Luo;Chunying Zhu;Lixin Hu;Meng Sun
Generative artificial intelligence (GAI) has become an epoch-making technology in the educational context. With a quasi-experimental repeated measure design and mixed-method data collection, this study examined the effects of the GAI-based chatbot in assisting preservice teachers in implementing the new national curriculum standards in Mainland China and their perceptions accordingly. A sample of 26 preservice teachers (divided into 13 teams) was included in this two-phrase study. Results showed that textbook design activities with the chatbot effectively promoted participants’ acquisition of content knowledge and improved self-efficacy, although it did not reduce teaching anxiety. Evidence was also extracted from participants’ open-ended responses with an extended COSTEM (i.e., content, others, self, tasks, ethics, and model) framework. Meanwhile, preservice teachers perceived both advantages and disadvantages regarding the utility of the GAI-based chatbot in learning. Implications of this study were also discussed.
{"title":"Empowering Preservice Teachers Through Textbook Design Activities With GAI-Based Chatbot","authors":"Jiutong Luo;Chunying Zhu;Lixin Hu;Meng Sun","doi":"10.1109/TLT.2025.3606757","DOIUrl":"https://doi.org/10.1109/TLT.2025.3606757","url":null,"abstract":"Generative artificial intelligence (GAI) has become an epoch-making technology in the educational context. With a quasi-experimental repeated measure design and mixed-method data collection, this study examined the effects of the GAI-based chatbot in assisting preservice teachers in implementing the new national curriculum standards in Mainland China and their perceptions accordingly. A sample of 26 preservice teachers (divided into 13 teams) was included in this two-phrase study. Results showed that textbook design activities with the chatbot effectively promoted participants’ acquisition of content knowledge and improved self-efficacy, although it did not reduce teaching anxiety. Evidence was also extracted from participants’ open-ended responses with an extended COSTEM (i.e., content, others, self, tasks, ethics, and model) framework. Meanwhile, preservice teachers perceived both advantages and disadvantages regarding the utility of the GAI-based chatbot in learning. Implications of this study were also discussed.","PeriodicalId":49191,"journal":{"name":"IEEE Transactions on Learning Technologies","volume":"18 ","pages":"822-832"},"PeriodicalIF":4.9,"publicationDate":"2025-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145141688","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01DOI: 10.1109/TLT.2025.3604616
John Chi-Kin Lee;Chris Dede;Minjuan Wang;Xuefan Li
This article proposes a novel framework for ethical human–artificial intelligence (AI) partnerships in education by integrating Eastern ethics (with Chinese ethics as an example), intelligence augmentation, and agentic AI design. Moving beyond the dominant Western paradigm, the study draws from Confucian and Daoist principles—such as relational trust, coagency, and moral cultivation—to envision AI as an ethical partner, not just a tool. It addresses two key questions: How can trust in AI be cultivated in education? and when can AI be ethically considered a collaborator? The authors introduce a triadic model combining normative grounding, cognitive scaffolding, and system-level design, operationalized through culturally sensitive platforms, pedagogy, and ethical interaction. They also propose a three-tiered evaluation system: learner trust metrics, educator audits, and AI reflexivity protocols. This interdisciplinary synthesis provides a scalable culturally rooted pathway for designing AI systems that are pedagogically meaningful, ethically adaptive, and co-constructive—contributing to more equitable and morally resonant educational futures.
{"title":"Building Trust in AI Through Dialogues With Eastern Ethics: Toward Ethical Partnerships in Education","authors":"John Chi-Kin Lee;Chris Dede;Minjuan Wang;Xuefan Li","doi":"10.1109/TLT.2025.3604616","DOIUrl":"https://doi.org/10.1109/TLT.2025.3604616","url":null,"abstract":"This article proposes a novel framework for ethical human–artificial intelligence (AI) partnerships in education by integrating Eastern ethics (with Chinese ethics as an example), intelligence augmentation, and agentic AI design. Moving beyond the dominant Western paradigm, the study draws from Confucian and Daoist principles—such as relational trust, coagency, and moral cultivation—to envision AI as an ethical partner, not just a tool. It addresses two key questions: How can trust in AI be cultivated in education? and when can AI be ethically considered a collaborator? The authors introduce a triadic model combining normative grounding, cognitive scaffolding, and system-level design, operationalized through culturally sensitive platforms, pedagogy, and ethical interaction. They also propose a three-tiered evaluation system: learner trust metrics, educator audits, and AI reflexivity protocols. This interdisciplinary synthesis provides a scalable culturally rooted pathway for designing AI systems that are pedagogically meaningful, ethically adaptive, and co-constructive—contributing to more equitable and morally resonant educational futures.","PeriodicalId":49191,"journal":{"name":"IEEE Transactions on Learning Technologies","volume":"18 ","pages":"833-841"},"PeriodicalIF":4.9,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145141691","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01DOI: 10.1109/TLT.2025.3604522
Chu-Fu Wang;Yih-Kai Lin;Ling Cheng
In testing systems, the item response theory is a widely used model for accurately synthesizing user response information. However, compared to classical test theory approaches, it imposes a higher computational burden and increases the system design complexity. Quantum computing has shown promise in alleviating these computational challenges. Currently, general-purpose quantum computers are still in a relatively early stage of development. However, special-purpose quantum computing architectures have been designed to solve combinatorial optimization problems, attracting significant attention across various fields. These systems enable researchers to tackle domain-specific optimization problems with reduced computational time. To the best of our knowledge, no applications of quantum computing have been proposed in the field of educational technology. This study, therefore, aimed to design a quantum quadratic unconstrained binary optimization formulation for optimizing test sheet composition. The proposed model can be implemented on practical quantum Ising machines (or digital quantum Ising machines for larger qubit usage) to evaluate system efficiency. Simulation results demonstrate that the proposed approach outperforms traditional methods, including the genetic algorithm and particle swarm optimization algorithm, in terms of computational efficiency.
{"title":"Quantum Algorithm Design and Its Implementation for Solving Test Sheet Composition Optimization Using a Quantum Annealing Approach","authors":"Chu-Fu Wang;Yih-Kai Lin;Ling Cheng","doi":"10.1109/TLT.2025.3604522","DOIUrl":"https://doi.org/10.1109/TLT.2025.3604522","url":null,"abstract":"In testing systems, the item response theory is a widely used model for accurately synthesizing user response information. However, compared to classical test theory approaches, it imposes a higher computational burden and increases the system design complexity. Quantum computing has shown promise in alleviating these computational challenges. Currently, general-purpose quantum computers are still in a relatively early stage of development. However, special-purpose quantum computing architectures have been designed to solve combinatorial optimization problems, attracting significant attention across various fields. These systems enable researchers to tackle domain-specific optimization problems with reduced computational time. To the best of our knowledge, no applications of quantum computing have been proposed in the field of educational technology. This study, therefore, aimed to design a quantum quadratic unconstrained binary optimization formulation for optimizing test sheet composition. The proposed model can be implemented on practical quantum Ising machines (or digital quantum Ising machines for larger qubit usage) to evaluate system efficiency. Simulation results demonstrate that the proposed approach outperforms traditional methods, including the genetic algorithm and particle swarm optimization algorithm, in terms of computational efficiency.","PeriodicalId":49191,"journal":{"name":"IEEE Transactions on Learning Technologies","volume":"18 ","pages":"842-855"},"PeriodicalIF":4.9,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145141690","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
As programming education scales, evaluating student code becomes increasingly challenging. In object-oriented programming (OOP) courses, design patterns are crucial for teaching maintainable reusable solutions that meet industry standards. While traditional automated assessment tools have successfully addressed code correctness and quality, automating the detection of design patterns presents a distinct challenge. Current methods, such as static code analysis combined with graph similarity, have proven effective for library code but often struggle with the variability of student submissions. This article investigates the application of generative artificial intelligence (GenAI) models to improve accuracy in detecting design patterns in student-written code. Our research addresses two key questions: 1) Which of the current GenAI models offer optimal performance, accuracy, and reasoning capabilities for design pattern assessment? and 2) How does a cloud-based Software as a Service solution (such as ChatGPT) compare to a cloudlet solution (local model deployed on the University’s cluster) in terms of reliability and scalability? We assess the effectiveness of these approaches using a representative sample of student assignments specifically crafted to require design pattern implementation. Our findings discuss the educational utility of GenAI in reducing instructors’ grading burdens, enhancing students’ self-assessment opportunities, and its potential to guide industry practitioners in design pattern evaluation. We highlight human-in-the-loop supervision to ensure that GenAI recommendations align with educational standards and advocate for open-source cloudlets as a scalable cost-effective infrastructure for universities. Our study provides guidelines for integrating GenAI into OOP coursework and highlights the implications for teaching, industry applicability, and future research directions.
{"title":"Using GenAI to Assess Design Patterns in Student-Written Code","authors":"Daniel-Florin Dosaru;Diana-Maria Simion;Andrei-Horia Ignat;Lorina-Cristina Negreanu;Alexandru-Corneliu Olteanu","doi":"10.1109/TLT.2025.3604054","DOIUrl":"https://doi.org/10.1109/TLT.2025.3604054","url":null,"abstract":"As programming education scales, evaluating student code becomes increasingly challenging. In object-oriented programming (OOP) courses, design patterns are crucial for teaching maintainable reusable solutions that meet industry standards. While traditional automated assessment tools have successfully addressed code correctness and quality, automating the detection of design patterns presents a distinct challenge. Current methods, such as static code analysis combined with graph similarity, have proven effective for library code but often struggle with the variability of student submissions. This article investigates the application of generative artificial intelligence (GenAI) models to improve accuracy in detecting design patterns in student-written code. Our research addresses two key questions: 1) Which of the current GenAI models offer optimal performance, accuracy, and reasoning capabilities for design pattern assessment? and 2) How does a cloud-based Software as a Service solution (such as ChatGPT) compare to a cloudlet solution (local model deployed on the University’s cluster) in terms of reliability and scalability? We assess the effectiveness of these approaches using a representative sample of student assignments specifically crafted to require design pattern implementation. Our findings discuss the educational utility of GenAI in reducing instructors’ grading burdens, enhancing students’ self-assessment opportunities, and its potential to guide industry practitioners in design pattern evaluation. We highlight human-in-the-loop supervision to ensure that GenAI recommendations align with educational standards and advocate for open-source cloudlets as a scalable cost-effective infrastructure for universities. Our study provides guidelines for integrating GenAI into OOP coursework and highlights the implications for teaching, industry applicability, and future research directions.","PeriodicalId":49191,"journal":{"name":"IEEE Transactions on Learning Technologies","volume":"18 ","pages":"869-876"},"PeriodicalIF":4.9,"publicationDate":"2025-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11144915","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145351916","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：