The rapid development of educational robots (ERs) and their introduction in various educational fields have made the topic of ERs increasingly important, especially in K-12 education. This study aims to provide educational technology researchers and teachers with a more comprehensive understanding of the research focus and the application of ERs in K-12 technology-assisted learning environments. Adopting a technology-based learning model and using the Web of Science (WoS) database as the primary data source, this study systematically reviewed the application issues, research issues, and interaction issues of ERs in K-12 (ER-K12) publications in education science and technology SSCI journals from 2005 to 2024. We further analyzed the ER roles, distribution of types, and learning strategies used in different application domains. The review results showed that ER-K12 research is growing rapidly, with science and programming as the main subjects, but research in other subjects, such as language and arts, is also increasing. After cross-comparison analysis, this study found that there are clear differences in the types of robot applications, roles of robots, and learning strategies adopted in different subjects using ERs. Science and programming prefer technology-oriented applications, where ERs are often used as tools with structured and inquiry-oriented strategies. Language and arts, on the other hand, emphasize interaction and creative expression, and ERs often take on the roles of instructors and peers, with contextual learning and digital storytelling. Based on the above findings, this study also suggests future research directions for researchers and teachers.
This study investigates the use of local Large Language Models (LLMs): Gemma, LLama, Mistral Small, and Phi, in assessing Spanish second language (L2) writing. While commercial models like ChatGPT have already been used for this task, they present privacy risks and reproducibility challenges. To evaluate local alternatives to ChatGPT, human raters were used to assess a third-semester assignment, evaluating 100 paragraphs to provide a baseline of comparison for the local LLMs’ assessment. Results showed varying performance: while some models approached the human baseline; others were inaccurate and inconsistent. On average, they performed better using a holistic rubric than an analytic one. A third experiment, averaging the top 3 models: Llama3.3:70b, Gemma2:9b, Mistral-Small:22b, achieved the highest agreement (r = 0.88), surpassing inter-human rating. A separate experiment testing RAG and fine-tuning techniques on the smallest model: Llama3.1:8b, showed no significant improvements with RAG but substantial gains in accuracy and consistency with fine tuning. Findings suggest that local LLMs show promise for Spanish L2 writing assessment. Local LLMs are preferable to proprietary ones because they protect privacy and reduce risks of data exposure, they may better support replicable workflows and may offer practical advantages for long term adoption.
With the rapid development of generative artificial intelligence (GenAI), leveraging its capabilities to optimize video learning has emerged as a critical research topic. However, empirical evidence about its effectiveness and underlying cognitive mechanisms remains scarce. This study recruited 120 undergraduate and postgraduate students and employed multimodal analysis to investigate the impact of a GenAI-supported questioning mechanism (AI-VL mode) on learners’ intrinsic motivation, cognitive load, attention allocation, and learning performance. Results showed that the AI-VL mode significantly outperformed traditional video learning (TVL) across intrinsic motivation, cognitive load, and learning performance, and further outperformed embedded question-and-answering video learning (QA-VL) specifically in learning performance. Regarding cognitive processing, eye-tracking metrics confirmed that the AI-VL mode redistributed cognitive resources through real-time feedback, reducing intrinsic and extraneous cognitive load. Furthermore, Lag Sequential Analysis of interactive behaviors revealed a cross-modal attentional cycle, specifically the “feedback-viewing” path, which enabled the reallocation of liberated cognitive resources into germane cognitive load. However, cluster analysis revealed that most learners remained at a surface-level interaction stage. These findings provide empirical evidence and practical guidance for designing adaptive GenAI-driven educational scaffolds for future online video learning platforms.
YouTube is one of the most widely used social media platforms, with millions of videos uploaded by thousands of people every day. Some of the uploaded videos can be used for educational purposes. However, there is no mechanism to test the reliability of the information in these videos. For this reason, in recent years, scientists have been studying YouTube video analysis. This study aims to conduct a bibliometric analysis to examine the scope of studies on YouTube video analysis and to reveal the trend of the subject. For this purpose, 2352 articles published in indexed journals between 2007 and 2023 in the Web of Science database were analyzed. In the analysis conducted using VOSviewer, R Studio, and CiteSpace, it was determined that there was a significant increase in the number of articles after 2019. It has been determined that the most used keywords in this field are YouTube and social media. However, upon examining the evolution of the subject, it was found that keywords shifted to concepts such as education and quality assessment over time. It was revealed that quality assessment scales (DISCERN, JAMA, and GQS) have been adopted as keywords, particularly since 2022. Furthermore, the authors, journals, and countries that use these keywords most frequently were identified. The country that contributed the most to the related field was the USA, and the country with which this country had the highest relationship was Turkey.
GenAI is reshaping critical thinking in higher education, even as it blurs the distinction between authentic intellectual effort and AI-assisted performance and complicates the design of assessments and the interpretation of learning outcomes. However, the extent to which the theoretical, methodological, and empirical findings of existing studies support these assumptions remains unclear. Therefore, this integrated bibliometric and systematic review of 39 peer-reviewed studies published between 2022 and 2024 was conducted to examine the development of GenAI–critical thinking research over time and identify the bibliometric patterns, the theoretical frameworks guiding this work, the methodological approaches most commonly used, and the reported effects of GenAI on students’ critical thinking skills. Findings suggest that, even with the rapidly expanding research on ChatGPT and critical thinking as core keywords in the United States, China, and the United Kingdom, and a marked rise in publications in 2023, the conceptual and methodological clarity of the impact of GenAI on critical thinking remains wanting. Of the 39 articles, 17 (44
This systematic review examines how AI-driven tools and innovative pedagogical models influence key outcomes in programming education, with a focus on student engagement, comprehension, retention, and self-regulation. A structured search was conducted in the Scopus database for peer-reviewed journal articles published between 2010 and 2023, and the selection process followed the PRISMA 2020 guidelines. After applying predefined inclusion and exclusion criteria, 29 empirical studies on flipped classrooms, gamification, project-based learning, collaborative learning, and blended learning in programming or computational-thinking courses were included in the review. Across these models, adaptive learning platforms, automated feedback systems, and collaborative tools were found to enhance active learning and support more personalized and self-regulated study. At the same time, the effectiveness of AI-enhanced learning environments depends on careful scaffolding, attention to accessibility, and a balanced use of motivational strategies. For educators, a core implication of the review is that combining AI-based adaptive feedback with structured in-class practice and well-designed collaborative activities can help reduce dropout, sustain motivation, and foster durable programming competencies. The findings offer evidence-informed guidance for designing programming courses that leverage AI and innovative pedagogies to create more engaging and inclusive learning experiences.
This study investigates the integration of Artificial Intelligence (AI) into university teaching through the lens of the Intelligent-TPACK framework, emphasizing the interplay between AI knowledge, pedagogical strategies, and ethical decision-making. Using a two-phase mixed-methods approach, we first quantitatively examined how university educators' AI proficiency influences their pedagogical and content knowledge. The results showed that while AI literacy strengthens instructional capacities, it does not automatically translate to a holistic Intelligent-TPACK framework unless supported by ethical and pedagogical components. The second phase involved qualitative interviews to explore how educators navigate AI-driven instruction, address ethical concerns, and make informed instructional decisions. The findings reveal that while educators appreciate AI’s potential for personalized learning and real-time feedback, they remain concerned about algorithmic bias, transparency, and over-reliance on automation. The study underscores the urgent need for structured AI literacy programs that integrate technological, pedagogical, and ethical training. By fostering interdisciplinary collaboration and establishing ethical AI guidelines, universities can ensure that AI serves as a pedagogical enabler rather than a replacement for human-centered instruction.
We investigate how students’ learning experiences and performance differ when engaging with an AI-generated instructional video versus a human-made video. We address two new aspects when developing the AI-generated video: (1) Students’ connection with the character of the AI-video, the character being the instructor of the course, minimizing thus potential disruptions and (2) the technical aspect of the content. The study followed a quasi-experimental, non-randomized matched-group design involving 69 students who were enrolled in a business mathematics course offered during their first year of study and were taught by the same professor. The participants were assigned to two groups: the experimental group where students watched an AI-generated instructional video, and a control group where students watched a human-made instructional video. The learning experience was measured using five dimensions, namely comprehension, learner engagement, pace, instructor presence and overall satisfaction. The learning performance was measured using a retention test and a transfer test. We found no significant differences in the learning performance between the two groups. However, we found significant differences between the groups on learner engagement with small to medium effect size and on instructor presence with medium effect size. This suggests that AI-generated videos may achieve comparable learning outcomes to traditional videos in specific, controlled contexts. Finally, we underscore the potential of AI-generated videos as scalable educational tools while noting the need for further refinement in areas like voice quality and for future research involving larger, more diverse samples.
Advances in artificial intelligence (AI) and machine learning (ML) are increasingly central to predicting student achievement in science and STEM education. This study uses a quantitative bibliometric design to map the conceptual, methodological, and thematic structure of research in this domain. The dataset comprises 1,073 English-language peer-reviewed journal articles indexed in the Web of Science (WoS) between 1993 and 2025, identified through PRISMA-informed screening and analyzed using Bibliometrix/Biblioshiny. Student achievement is operationalized via outcome-oriented constructs reflected in article metadata (e.g., grades, test performance, persistence/retention, and dropout risk), rather than through primary data synthesis. Findings show strong publication growth after 2010 and a recent decline in average citation impact, largely attributable to shorter citation windows for newer publications. The United States and China lead publication output, while Saudi Arabia, the United Kingdom, and Mexico show high international collaboration. Purdue University and Texas A M University are among the most productive institutions. IEEE Access, Computers Education, and Education and Information Technologies emerge as core outlets, and D. Gasevic is identified as the most influential author. Thematic and co-word analyses indicate three dominant knowledge clusters: AI/ML and predictive modelling methods, science/STEM learning contexts, and student-related factors, alongside an emerging axis focused on ethics, explainability, and large language models (LLMs). Overall, the map clarifies dominant themes, evolving trends, and gaps that can inform future comparative studies and responsible AI integration. Limitations include reliance on a single database (WoS) and English-only publications, which may constrain the representativeness of the findings.
Collaborative creativity (cocreativity) and Artificial Intelligence (AI) have been integrated into organisations to solve complex challenges. Simultaneously, AI has shown the potential to improve human cocreativity. However, a research gap persists related to the practical integration of AI in the cocreative process to enhance students’ cocreativity in the classroom environment. This paper studies real-time novelty evaluation using AI techniques, how to deliver real-time novelty feedback in a cocreative working space and explores its impact on the cocreation process. A case study-mixed research was designed, involving twenty undergraduate students working in groups to generate a novel solution to an open-ended science challenge using a Google Slide workspace. The co-generated ideas’ novelty was evaluated using the Universal Sentence Encoder technique, and students received real-time feedback on the novelty of their cocreated ideas. Descriptive statistics and qualitative analysis were performed to investigate the impact of receiving feedback on the level of novelty and the presence of cocreative behaviours. Real-time novelty feedback procedure can be successfully implemented during the cocreative process in Google Slides. Besides, the results show that the feedback enhances novelty in the ideas. Furthermore, the feedback received engages and promotes students’ cocreative behaviours, such as stopping when receiving feedback, reading, discussing, and applying it to their ideas. Real-time feedback can be a valuable tool to promote the behaviours that can lead a group to generate novel solutions to tackle key 21st-century challenges. Moreover, it can have a strong pedagogical impact on learning and teaching methodologies for cocreativity.
Enhancing cybersecurity awareness and digital literacy in the Internet era is increasingly important for young people. Digital games, with their interactivity and immersion, have been applied to the cultivation of cybersecurity awareness among adolescents. However, general digital games lack targeted guidance, making it difficult to sustain deep engagement and limiting academic performance. To this end, this study proposes an AI Agent-Integrated Digital Game Learning Approach (AIDG-approach) that integrates an AI agent-based adaptive feedback system to enhance middle school students’ academic performance, learning motivation, and classroom engagement in the Principles of the Internet course. To verify the effectiveness of the approach, 102 students (mean age 12.3 years) from a middle school in eastern China were recruited and assigned to an experimental group (n = 53) and a control group (n = 49). The experimental group used a digital game learning approach (AIDG-approach) integrating an AI agent-based adaptive feedback system, and the control group used a conventional digital game learning approach (CDG-approach). The results of the study showed that the experimental group was better than the control group in terms of academic performance and learning motivation, but the difference was not significant; while in terms of classroom engagement, the experimental group had a significant improvement over the control group. The interview results further revealed that the AI agent’s personalized feedback feature effectively enhanced students’ learning interest and task engagement, but the accuracy of its feedback and task complexity still limited the further improvement of motivation and performance. This study provides empirical evidence for the application of an AI agent–based adaptive gamified learning approach, enriches relevant theoretical frameworks, and offers guidance for design and deployment in educational practice.
While educational gamification successfully drives student engagement, it faces persistent criticism for fostering extrinsic reward dependency, superficial achievement (“fast leveling”), and inequitable learning experiences. This study posits Artificial Intelligence (AI) as a critical corrective mechanism to these structural limitations, repositioning static game mechanics within dynamic, adaptive learning ecosystems. Adopting an Integrative Review methodology based on Whittemore and Knafl’s framework, this study synthesizes 61 empirical and theoretical studies (2003–2025) identified through systematic two-way snowballing. Complementary bibliometric data from Scopus and ScienceDirect reveals an exponential “J-curve” growth in the field, marking a decisive disciplinary shift from computer science architectures to pedagogical applications in Social Sciences. The findings indicate that AI integration mitigates traditional gamification pitfalls by (1) personalizing difficulty through adaptive algorithms, (2) replacing superficial rewards with intelligent, real-time feedback, and (3) enhancing inclusivity for diverse learner profiles. Crucially, this review proposes the “AI Corrective Role Framework,” a conceptual model grounded in convergent evidence that operationalizes how AI acts as a learner-centered function to deepen cognitive retention and as a decision-making instrument for institutional strategy. These insights offer researchers and policymakers a robust roadmap for implementing sustainable, evidence-based, and equitable gamified learning environments in the era of Generative AI.
This study investigates targeted professional development practices for online educators in higher education, moving beyond the conventional one-size-fits-all model. The goal is to determine effective strategies in teaching, community involvement, and institutional support to craft a comprehensive framework that boosts instructors’ performance. Interviews conducted with 13 accomplished faculty members from a prominent U.S. institution recognized for its excellence in online education revealed key insights. The findings underscore the critical importance of a holistic approach that integrates teaching expertise, community connections, and robust organizational backing. This Integrated triad model enhances the quality and effectiveness of online education, demonstrating that these intertwined supports magnify each other’s impact. With online education’s role becoming increasingly vital, these results are positioned to address a significant need and offer guidance for faculty members seeking to refine their online teaching capabilities, fostering better engagement and success in virtual classrooms.
In this systematic review, we examine how technological tools have enhanced collaborative language learning over the past decade (2014–2024). Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, we analyzed 52 empirical studies to identify technological tools, implementation approaches, and effectiveness patterns in technology-enhanced collaborative language learning (TECLL). Our analysis identified a variety of technologies, grouped into synchronous tools (social media, video conferencing tools, and gamified platforms) and asynchronous tools (cloud-based platforms and learning management systems). These technologies support distinct collaborative language-learning processes through different affordances. Synchronous tools facilitate the immediate negotiation of meaning and real-time interactions, particularly enhancing speaking, listening, and pragmatic competence. Asynchronous tools support extended reflection and recursive revision, and have been shown to have a significant impact on writing development and metalinguistic awareness. Effective implementations across technological categories share common characteristics, such as structured collaboration processes, authentic communicative purposes, and balanced teacher presence. This review contributes to the field in three key ways: (1) offering a research-based classification of technology tools for collaborative language learning, (2) summarizing implementation strategies that apply across technologies, and (3) highlighting effectiveness trends to support evidence-based teaching. These results offer teachers practical, evidence-based advice for choosing and using technology to improve collaborative language learning in connected educational settings.
While the crucial role of teacher readiness in technology-assisted teaching and learning has been well documented, studies on K-12 language teachers’ teaching readiness in the flipped classroom remain relatively few. Informed by the technological pedagogical content knowledge (TPACK) and technology acceptance (TAM) theories, the current research examined K-12 language teachers’ flipped teaching readiness and its relationships with perceived anxiety of flipped teaching and teacher self-efficacy. Data were drawn from 489 K-12 pre-service and in-service Chinese language teachers. The findings showed that teacher participants reported a generally high level of flipped language teaching readiness, with gender differences only existing among pre-service language teachers’ flipped teaching readiness. Teacher self-efficacy was found to be significantly related to both pre-service and in-service language teachers’ flipped language teaching readiness, while perceived anxiety of flipped teaching was not associated with flipped language teaching readiness. The findings of this research shed light on how K-12 language teachers’ flipped teaching readiness can be enhanced in language education.
Computational thinking (CT) is an essential 21st -century competency that can be learned in early childhood. While both plugged (digital) and unplugged (non-digital) programming activities are advocated for developing CT in young children, their comparative effectiveness remains inadequately explored. This quasi-experimental study addressed this gap by evaluating the impact of plugged versus unplugged programming on CT skills and programming interest in 104 5-6-year-old kindergarteners from China with no prior exposure. Compared to a business-as-usual control group, both experimental interventions significantly enhanced CT. Crucially, children in the plugged programming environment demonstrated significantly greater gains in CT skills and programming interest than those in the unplugged group. These findings carry important implications for educational practice and policy in the AI era: they provide evidence-based guidance for selecting tools, suggest that plugged approaches offer distinct advantages for CT development, and inform decisions about resource allocation and technology integration in early childhood education. The study further advances our theoretical understanding by identifying feedback mechanisms as a key factor explaining the differential outcomes.
In the current trend of generative artificial intelligence (GAI) as an assistive tool for teaching and learning, it is crucial for teachers to employ GAI for teaching and to guide students in knowledge construction. The present study adopted the draw-a-picture technique and epistemic network analysis (ENA) to reveal pre-service teachers’ (PSTs) perspectives on GAI-assisted instruction. It focused on examining the differences in the conceptions between realistic and idealistic GAI-assisted instruction, as well as the differences in realistic and idealistic GAI-assisted instruction among PSTs with varying levels of learning attitudes. The results showed that in the conceptions of realistic GAI-assisted instruction, the majority of PSTs tended to depict individual learners learning through screens. Conversely, in the conceptions of idealistic GAI-assisted instruction, they tended to illustrate multiple learners participating in learning activities under the direct guidance of a teacher. In terms of the categories of “locations,” “activities,” and “learning content,” most PSTs held consistent viewpoints between the realistic and idealistic conceptions, while differences existed in the categories of “participants involved,” “objects,” and “emotions and attitudes.” Furthermore, PSTs with high-level learning attitudes did not mention specific subjects but emphasized positive emotions, learning activities, and individual students. Conversely, PSTs with low-level learning attitudes held conservative attitudes toward the innovative application of GAI in education, preferring conventional educational modes and familiar teaching tools. This study not only contributes to understanding PSTs’ perspectives on GAI-assisted instruction, but also provides recommendations for future training related to GAI-assisted instruction of PSTs.
Inquiry-based learning has been a promising method in science curricula and many research projects. However, learners often struggle with low scientific inquiry skills, performance, and self-efficacy. To address these problems, this study aims to facilitate inquiry-based learning through generative artificial intelligence (GenAI) technologies. The current study compared three approaches to facilitating inquiry-based learning. In total, 120 pupils from three classes in Grade 5 participated in this study. Each was assigned to either experimental group 1, which used GenAI-based pedagogical agents to carry out inquiry-based learning, experimental group 2, which used GenAI-based chatbots to conduct inquiry-based learning, or the control group, which used a traditional learning platform to conduct inquiry-based learning. The results indicated that the knowledge gains, scientific inquiry skills, inquiry-based learning performance, and self-efficacy of the two experimental groups were substantially greater than those of the control group. There were no significant differences in knowledge gains, scientific inquiry skills, inquiry-based learning performance, or self-efficacy between experimental group 1 and 2. This study made two core contributions. First, it elucidates the significant influence of GenAI on scientific inquiry skills and inquiry-based learning performance. Second, it examines actionable strategies for the effective integration of GenAI within inquiry-based learning. Therefore, the present study advocates for the application of GenAI as a transformative tool to advance future inquiry-based learning practice.
The recent introduction of Computational Thinking (CT) in K-6 education has yet to clarify the corpus of content and competencies to be developed in these early stages. Numerous studies address CT education through different classroom approaches. This scoping review examines the literature, instructional models, materials, activities, and resources employed in K-6 classrooms to integrate CT into primary education. Findings suggest a certain misalignment between the broad definition of CT proposed in theoretical frameworks and the predominant focus on coding and programming in classroom practices. Although numerous CT components are extracted from the theoretical frameworks in the literature, classroom activities tend to work on a narrower set, with algorithmic thinking being the only focus in many of them. Programming practices range from digital storytelling to robotics, while a few include unplugged or interdisciplinary activities. Mathematics emerge as the most frequent course where CT is integrated, largely due to their natural overlap. Other courses such as science, language, and the arts also offer a significant potential for its appearance. In terms of instructional models, interactive, student-centered approaches such as guided discovery and problem-based learning are predominant. This review serves as a critical reflection on the alignment between the theoretical aspirations for CT integration and the practical realities of classroom implementation, shedding light on gaps and opportunities for refinement in pedagogical practices.
Learner emotions in collaborative learning have been actively researched. This paper collected 1866 papers about learner emotions in collaborative learning contexts from 2000 to 2023 and analysed them in terms of (1) research topics, (2) topics’ evolutional patterns, (3) contributing countries/regions/institutions, and (4) scientific collaborations using topic modelling and bibliometrics. Our primary emphasis was on analysing article patterns, discerning the involvement of institutions, countries, or regions, illustrating partnerships via social network visualization, and uncovering significant themes and their developmental shifts via topic modelling and statistical analysis. Results indicated a steady increase in articles focusing on emotions in collaborative learning, attributed largely to the substantial input from scholars based in the USA, the UK, and Australia, alongside a rising interest among academics from Spain, China, as well as Taiwan, Province of China. Examination of regional and institutional partnerships uncovered a tendency for closely located institutions to engage in collaborative endeavours, particularly when sharing similar research focuses. The analysis of topics highlighted that learners’ emotions in collaborative learning, particularly online/computer-supported collaborative learning in various settings (e.g., science, teacher, healthcare, clinical education, and language education) received increasing attention. Diverse applications (e.g., digital games, Wiki, and digital learning platforms) have flourished into collaborative learning activities to understand learners’ behaviour patterns and emotional dynamics. Based on the results, a conceptual framework including seven key dimensions (i.e., contexts, emotional dimensions, collaborative methods and approaches, moderating and mediating variables, outcomes, subject domains, and theoretical foundations) was proposed to guide research on emotional experiences in collaborative learning.
