Computers & Education最新文献

{"title":"Towards reliable generative AI-driven scaffolding: Reducing hallucinations and enhancing quality in self-regulated learning support","authors":"Keyang Qian ,&nbsp;Shiqi Liu ,&nbsp;Tongguang Li ,&nbsp;Mladen Raković ,&nbsp;Xinyu Li ,&nbsp;Rui Guan ,&nbsp;Inge Molenaar ,&nbsp;Sadia Nawaz ,&nbsp;Zachari Swiecki ,&nbsp;Lixiang Yan ,&nbsp;Dragan Gašević","doi":"10.1016/j.compedu.2025.105448","DOIUrl":"10.1016/j.compedu.2025.105448","url":null,"abstract":"<div><div>Generative Artificial Intelligence (GenAI) holds a potential to advance existing educational technologies with capabilities to automatically generate personalised scaffolds that support students’ self-regulated learning (SRL). While advancements in large language models (LLMs) promise improvements in the adaptability and quality of educational technologies for SRL, there remain concerns about the hallucinations in content generated by LLMs, which can compromise both the learning experience and ethical standards. To address these challenges, we proposed GenAI-enabled approaches for evaluating personalised SRL scaffolds before they are presented to students, aiming for reducing hallucinations and improving overall quality of LLM-generated personalised scaffolds. Specifically, two approaches are investigated. The first approach involved developing a multi-agent system approach for reliability evaluation to assess the extent to which LLM-generated scaffolds accurately target relevant SRL processes. The second approach utilised the “LLM-as-a-Judge” technique for quality evaluation that evaluates LLM-generated scaffolds for their helpfulness in supporting students. We constructed evaluation datasets, and compared our results with single-agent LLM systems and machine learning approach baselines. Our findings indicate that the reliability evaluation approach is highly effective and outperforms the baselines, showing almost perfect alignment with human experts’ evaluations. Moreover, both proposed evaluation approaches can be harnessed to effectively reduce hallucinations. Additionally, we identified and discussed bias limitations of the “LLM-as-a-Judge” technique in evaluating LLM-generated scaffolds. We suggest incorporating these approaches into GenAI-powered personalised SRL scaffolding systems to mitigate hallucination issues and improve the overall scaffolding quality.</div></div>","PeriodicalId":10568,"journal":{"name":"Computers & Education","volume":"240 ","pages":"Article 105448"},"PeriodicalIF":10.5,"publicationDate":"2025-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145021041","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Digital innovations in teacher recruitment: An experimental study","authors":"Robert M. Klassen ,&nbsp;Hui Wang ,&nbsp;Joe Cutting ,&nbsp;Sophie Thompson-Lee ,&nbsp;Rebecca J.S. Snell ,&nbsp;Beng Huat See ,&nbsp;Michael Saiger","doi":"10.1016/j.compedu.2025.105450","DOIUrl":"10.1016/j.compedu.2025.105450","url":null,"abstract":"<div><div>Teacher shortages are a serious challenge in many countries, made worse by declining enrolments in initial teacher education (ITE) programs and growing competition for talented young people making career decisions. To address this challenge, we developed and tested two digital interventions—a persuasive game (TeachQuest) and a realistic job preview (RJP)—designed to enhance undergraduate students' teaching interest, teaching self-efficacy, and perceptions of fit with the profession. In a two-phase experimental study (<em>N</em> = 957), undergraduate participants were randomly assigned to TeachQuest, the RJP, or a control condition. Results from Phase 1 showed that both interventions increased participants' interest and perceived fit with teaching, with the RJP also improving teaching self-efficacy. In Phase 2, results from a delayed post-test (six weeks later; <em>N</em> = 572) indicated that while motivation-related outcomes remained higher than pre-test levels, changes were non-linear, with TeachQuest sustaining interest through participants’ immersion experiences and the RJP maintaining self-efficacy through mastery experiences. Our findings suggest that immersive game-based recruitment interventions may be particularly effective in informing and engaging potential applicants, whereas RJPs may be useful in reinforcing confidence in teaching. The study provides new insights informing the design of scalable, evidence-based teacher recruitment tools that align with the interests and digital-focused lives of prospective applicants for ITE programs.</div></div>","PeriodicalId":10568,"journal":{"name":"Computers & Education","volume":"239 ","pages":"Article 105450"},"PeriodicalIF":10.5,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144996720","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enhancing responsive teaching through in-the-moment interpretations of student resources: A study in AI-supported virtual simulation","authors":"Nuodi Zhang ,&nbsp;Fengfeng ke ,&nbsp;Chih-Pu Dai ,&nbsp;Alex Barrett ,&nbsp;Saptarshi Bhowmik ,&nbsp;Sherry A. Southerland ,&nbsp;Luke A. West ,&nbsp;Xin Yuan","doi":"10.1016/j.compedu.2025.105449","DOIUrl":"10.1016/j.compedu.2025.105449","url":null,"abstract":"<div><div>Responsive teaching, a pedagogical approach that foregrounds and builds instruction on student ideas, requires teachers to attend to and build on student resources. However, teachers' interpretations of student resources, especially during live teaching, remain understudied. In this study, we examined <em>in-the-moment interpretations</em>, teachers' real-time sense-making of and reflection on students' epistemic and emotional resources, and explored how teachers' in-the-moment interpretations can support their responsive teaching talk moves and knowledge. Employing a convergent mixed-methods research design, we designed and implemented a generative artificial intelligence (AI)-supported virtual simulation as a pedagogical sandbox for 40 preservice teachers (PSTs) to practice teaching with virtual students, interpret student resources, and act on these interpretations in real time. Linear regression analysis was conducted and found that PSTs’ in-the-moment interpretations are significant predictors of their responsive teaching talk moves and knowledge. Qualitative thematic analysis identified themes that corroborated and extended the findings of the quantitative component. Implications for teacher education and simulation design are discussed.</div></div>","PeriodicalId":10568,"journal":{"name":"Computers & Education","volume":"239 ","pages":"Article 105449"},"PeriodicalIF":10.5,"publicationDate":"2025-08-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144922494","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"How children learn from robots: Educational implications of communicative style and gender in child–robot interaction","authors":"Konrad Maj,&nbsp;Ariadna Gołębicka,&nbsp;Zuzanna Siwińska","doi":"10.1016/j.compedu.2025.105445","DOIUrl":"10.1016/j.compedu.2025.105445","url":null,"abstract":"<div><div>This study examined how primary school children respond to the communicative style and perceived gender of a humanoid robot during a controlled educational interaction. A total of 251 children (ages 7–12) interacted with a robot programmed to use either a polite asking style or a firm commanding style, and presented with either a female or male persona. We investigated whether children would imitate the robot's style (H1), and whether child age and gender would predict their tendency to anthropomorphize the robot (H2–H3). Results showed that children interacting with a polite robot almost always responded politely, whereas those encountering a commanding robot still overwhelmingly responded in a polite manner rather than mirroring its tone. Younger children and girls displayed significantly higher levels of anthropomorphization of the robot. Contrary to expectations (H4), the degree of imitation did not correlate with anthropomorphism. An ANOVA (H5) indicated that the robot's persona (gender × communication style) influenced anthropomorphism: the polite-female robot elicited the highest anthropomorphism scores, though post-hoc differences were nonsignificant. A regression analysis (H6) confirmed child age and gender as significant predictors of anthropomorphization. These findings underscore the importance of social cues in child–robot educational interactions. Tailoring a robot's communication style to children's developmental level and social expectations can enhance children's engagement and potentially support positive learning outcomes.</div></div>","PeriodicalId":10568,"journal":{"name":"Computers & Education","volume":"239 ","pages":"Article 105445"},"PeriodicalIF":10.5,"publicationDate":"2025-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144925131","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A systematic reexamination of field experiences in PK-12 online environments in the United States","authors":"Lauren J. Woo ,&nbsp;Leanna Archambault ,&nbsp;Jered Borup ,&nbsp;Ray R. Buss ,&nbsp;Danah Henriksen","doi":"10.1016/j.compedu.2025.105447","DOIUrl":"10.1016/j.compedu.2025.105447","url":null,"abstract":"","PeriodicalId":10568,"journal":{"name":"Computers & Education","volume":"239 ","pages":"Article 105447"},"PeriodicalIF":10.5,"publicationDate":"2025-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144918877","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Transfer of self-efficacy: ICT self-efficacy and reading self-efficacy mediate the effect of ICT use on reading achievement","authors":"Chenlian Zhang ,&nbsp;Yiu-Kei Tsang ,&nbsp;Jinxin Zhu","doi":"10.1016/j.compedu.2025.105446","DOIUrl":"10.1016/j.compedu.2025.105446","url":null,"abstract":"<div><div>Past studies have explored the non-linear relationship between ICT use and reading achievement, primarily focusing on secondary school students. However, there is a lack of research examining this relationship among primary school students, who are at a critical stage in their reading development. Furthermore, the crucial transfer mechanism from students' ICT self-efficacy to their reading self-efficacy has been overlooked in the relationship between ICT use and reading achievement. Inspired by Bandura's theory of self-efficacy, this research investigated the non-linear associations between ICT use and reading achievement, as well as the transfer of self-efficacy beliefs. A large, representative sample of 3830 Hong Kong fourth-grade students from PILRS 2021 was used in the analyses. The results indicated that students who used ICT to find and read information for 30 min or less demonstrated higher reading achievement than those who did not use ICT. However, using ICT for over 30 min showed no direct association with reading achievement compared to the shorter usage. In contrast, for preparing reports and presentations using ICT, students using ICT for 30 min or less had no significant impact on reading achievement compared to non-users, while those using ICT for over 30 min showed improved reading achievement compared to the shorter usage. Our findings extend Bandura's theory into the digital era, underscoring the transfer of students' self-efficacy from the ICT domain to the reading domain.</div></div>","PeriodicalId":10568,"journal":{"name":"Computers & Education","volume":"239 ","pages":"Article 105446"},"PeriodicalIF":10.5,"publicationDate":"2025-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144918875","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The more the merrier? Examining the effects of a conversational agent on EFL learners’ speaking in three conditions","authors":"Yao Ma,&nbsp;Zhuo Wang,&nbsp;Hui Pang","doi":"10.1016/j.compedu.2025.105442","DOIUrl":"10.1016/j.compedu.2025.105442","url":null,"abstract":"<div><div>The present study innovatively investigates how a generative AI conversational agent (TalkFriend) impacts EFL university students’ oral English proficiency, considering both cognitive and affective dimensions. Forty-five students were randomly assigned to individual or paired learning (with ‘Lead’ and ‘Assisting’ roles interacting with TalkFriend). Employing a multi-modal approach, we uniquely integrated EEG brainwave data with oral tests, questionnaires, and interviews. Findings revealed significant overall proficiency gains. Notably, paired learning fostered superior improvements in communicative confidence and fluency compared to individual learning, which primarily saw fluency gains. Lead learners in paired settings also exhibited markedly higher learning interest, a factor significantly correlating with their neural activity (EEG). Pronunciation accuracy appeared to develop independently. Interpreted through Vygotsky's Zone of Proximal Development (ZPD), these findings inform a proposed four-quadrant ‘emotional ZPD’ conceptual model, highlighting the crucial interplay of cognitive, affective, and social support (from both AI and peers). Our research offers critical neurocognitive and socio-interactional insights for optimizing AI tools in language education.</div></div>","PeriodicalId":10568,"journal":{"name":"Computers & Education","volume":"239 ","pages":"Article 105442"},"PeriodicalIF":10.5,"publicationDate":"2025-08-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144913723","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Human-GenAI interaction for active learning in STEM education: State-of-the-art and future directions","authors":"Sofie Otto ,&nbsp;Rea Lavi ,&nbsp;Lykke Brogaard Bertel","doi":"10.1016/j.compedu.2025.105444","DOIUrl":"10.1016/j.compedu.2025.105444","url":null,"abstract":"<div><div>This systematic state-of-the-art review synthesizes findings from 50 studies examining the integration of GenAI into active learning models (such as problem-based learning, collaborative learning, and inquiry-based learning) within STEM education from high school to graduate levels. The analysis identifies five overarching categories of human–GenAI interaction: Tutoring, Co-creating, Processing, Coaching, and Simulating, primarily leveraged to support individual learners in developing problem-solving, critical thinking, and computational thinking skills. While the findings highlight GenAI's potential to support constructivist active learning, its application remains largely individual in scope. Moreover, challenges related to algorithmic bias, information reliability, privacy, and limited domain specificity constrain the orchestration of synergistic human-GenAI interaction, placing significant pedagogical demands on both educators and learners when interacting with GenAI-powered applications. Future research should explore how human-GenAI interactions can be orchestrated to support more active, collaborative, and context-sensitive learning environments. This includes supporting students in developing the competencies necessary to engage, individually and collaboratively, with GenAI tools reflectively, purposefully, and meaningfully in ways that enhance active learning.</div></div>","PeriodicalId":10568,"journal":{"name":"Computers & Education","volume":"239 ","pages":"Article 105444"},"PeriodicalIF":10.5,"publicationDate":"2025-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144918876","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Webcam Use and Its Role in Children’s Engagement and Achievement During Extended Remote Learning","authors":". Sharlyn Ferguson-Johnson Ph.D., . Allison M. Ryan Ph.D., . Kai S. Cortina Ph.D.","doi":"10.1016/j.compedu.2025.105443","DOIUrl":"https://doi.org/10.1016/j.compedu.2025.105443","url":null,"abstract":"Online classes substitute in-person instruction in K-12 general education, yet whether children’s webcam use in online classes is meaningful for their learning remains poorly understood. This study examined children’s webcam use and academic outcomes during the COVID-19 pandemic using data from 1,426 elementary students (49% male, 47% female, 93% White) across 65 classrooms, exploring how individual- and classroom-level webcam use, socioeconomic and demographic factors, and remote learning duration explain variation in children’s engagement and achievement. In May 2021, students reported their remote learning routines and online class participation since May 2020; responses were then linked with 2020-2021 MSTEP achievement and demographic data. Correlational analyses indicated individual webcam use varied as a function of children’s perceived engagement, academic achievement, socioeconomic disadvantage, and gender identity, while classrooms differed in mean webcam use based on socioeconomic and demographic composition, remote learning duration, and mean engagement and achievement. MANOVA results indicated classroom webcam norms had a larger effect on engagement during remote learning relative to individual use. Specifically, children in classrooms with <ce:italic>universal webcam-on norms</ce:italic> reported significantly higher mean emotional and behavioral engagement compared to children in classrooms with <ce:italic>moderate/mixed</ce:italic> or <ce:italic>low webcam-on norms.</ce:italic> Multilevel models revealed classmates’ webcam use explained a significant portion of the variance in children’s 2020–2021 achievement and was a stronger predictor of 2020-2021 achievement than individual webcam use. Overall, findings underscore the need to consider classroom-level webcam use policies and incentives when developing and refining effective online classroom structure and curricula for elementary school student populations.","PeriodicalId":10568,"journal":{"name":"Computers & Education","volume":"9 1","pages":""},"PeriodicalIF":12.0,"publicationDate":"2025-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144900521","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enhancing goal attainment in higher education with a scripted conversational agent: Effects of monitoring and reflection support in digital learning","authors":"Gabrielle Martins van Jaarsveld ,&nbsp;Jacqueline Wong ,&nbsp;Martine Baars ,&nbsp;Marcus Specht ,&nbsp;Fred Paas","doi":"10.1016/j.compedu.2025.105441","DOIUrl":"10.1016/j.compedu.2025.105441","url":null,"abstract":"<div><div>Self-regulated learning (SRL) is essential for academic success in higher education, yet many students struggle to effectively regulate their own learning behaviours. While goal-setting interventions can help students set high-quality goals as the foundation for their learning behaviours, additional supports are needed to help transition from goal setting into effective goal striving. This study examines the impact of monitoring and reflection supports, delivered through a scripted conversational agent, on students’ goal attainment, SRL skills, and academic performance. In this study, 84 undergraduate students were randomly assigned to one of four experimental conditions: Control, Monitoring Only, Reflection Only, or Monitoring &amp; Reflection. Over a five-week intervention, participants engaged in weekly goal-setting activities, with additional monitoring and/or reflection prompts depending on their assigned condition. Results showed that participants in the Monitoring Only and Monitoring &amp; Reflection conditions reported significantly higher goal attainment than those in the Reflection Only and Control groups, suggesting that monitoring plays a critical role in reinforcing goal-directed behaviour. While SRL skills improved across all conditions, no significant differences were found between groups, indicating that consistent goal setting alone may support SRL development. There were no significant effects of the intervention on academic performance. These findings highlight the immediate effectiveness of progress-monitoring activities and suggest that reflection may require longer intervention periods to have significant effects. This study supports the use of conversational agents for delivering scalable SRL support and provides insights into the design of multi-phase SRL supports in digital education.</div></div>","PeriodicalId":10568,"journal":{"name":"Computers & Education","volume":"239 ","pages":"Article 105441"},"PeriodicalIF":10.5,"publicationDate":"2025-08-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144887143","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}