Massive open online courses (MOOCs) can disseminate academic knowledge, such as complete courses, to students and lifelong learners from all backgrounds and anywhere. Using a quantitative approach, this study examines the relationship between engagement, acceptance, and use of blended MOOCs among 1875 STEM students at the University of Cape Coast (UCC), Ghana. The theoretical basis is the proposed revised Community of Inquiry (RCoI) framework and the Unified Theory of Acceptance and Use of Technology (UTAUT). A structured questionnaire was designed and distributed via Google Forms to a randomly selected sample of 3,506 students using a two-stage cluster sampling technique. The survey achieved an 82% response rate, with 2,875 participants completing it. Out of these 2,875 participants, 1,875 were STEM students, and they were explicitly used for this paper. Structural equation modelling was employed to analyse how the RCoI model—encompassing cognitive, social, teaching, and learning presences—affects students’ acceptance and use of blended MOOCs. According to the study, STEM students’ engagement significantly impacts their ability to utilise blended MOOCs. The findings showed that within blended MOOCs, the presence of RCoI had a significant positive relationship with STEM students’ engagement. Again, the findings reveal significant influences of STEM students’ engagement on performance expectancy, effort expectancy, social influence, and facilitating conditions, highlighting the need for administrative and technical support to enhance the effective adoption of blended MOOCs. This study contributes robust insights into the dynamics of student interaction and learning outcomes in blended learning environments such as blended MOOCs within developing countries.
Digitalization of textbooks constitutes a pivotal aspect of the digital transformation of education, playing an indispensable role in the development of a modern education system. Currently, the construction of digital textbooks in higher education lacks a shareable and reusable model. Crowdsourcing is an innovative concept that leverages collective wisdom to improve the quality of resources, transcending organizational boundaries and facilitating the flow of knowledge. This study adopts the design-based research paradigm and comprehensively applies a variety of research methods to construct and implement a crowdsourcing-based construction model for digital textbooks in higher education. The model includes a knowledge subsystem, a process subsystem, and an organizational subsystem, covering crowdsourcing participants such as professionals, ordinary participants and artificial intelligence. The digital textbook developed based on this framework has demonstrated significant user satisfaction in terms of content, teaching, technical design, evaluation, opportunities for deep learning, accessibility, and security. It is expected that the construction model of the digital textbooks in high education built in this study can further enrich the research outcomes of digital textbooks, and provide a model reference for digital textbook construction in higher education.
Robot programming has become an effective means to promote primary school students’ computational thinking (CT). Direct instruction (DI), characterized by teacher instruction followed by student problem-solving, is a commonly adopted teaching approach for robot programming and CT. However, by emphasizing task completion and minimizing failures, DI limits students’ ability to understand the essence of problem-solving and learn from mistakes. In contrast, productive failure (PF) involves students’ problem-solving followed by teacher instruction, treating failure as a valuable experience that fosters deeper learning and problem-solving skills. Hence, this research compared the effects of the productive failure-based robot programming (PF-RP) and direct instruction-based robot programming (DI-RP) approaches on primary school students’ CT, self-efficacy, growth mindset, and learning perceptions. A quasi-experimental study involving both quantitative and qualitative data was designed. The sample comprised 87 third-grade students from China, with 43 students in the PF-RP group and 44 students in the DI-RP group. Results revealed that (1) both approaches enhanced students’ CT, with the PF-RP approach demonstrating greater improvements than the DI-RP approach over time; (2) both approaches improved students’ self-efficacy, but no significant differences were detected between the two approaches over time; (3) only the PF-RP approach significantly promoted students’ growth mindset and showed greater improvements than the DI-RP approach over time; and (4) students perceived that the PF-RP approach is an effective teaching approach that helped them think of more solutions and strengthen their understanding.
In the age of artificial intelligence, empathy is crucial for preschoolers’ social adaptation and personal development. The development of Humanoid Intelligent Robots (HIRs) offers an opportunity to provide preschoolers with personalized interactive experiences. This study recruited 77 rural preschoolers aged 5 to 6 years as participants, who were randomly assigned to an R-tutor group, an R-peer group, and a conventional teaching group. An 8-week empathy course was conducted to explore the impact of different roles of HIRs on their empathy. The results of the questionnaire analysis showed that the R-tutor method improved preschoolers’ empathy more than the R-peer method and the conventional teaching method; the R-peer method did not enhance preschoolers’ empathy compared to the conventional teaching method. In addition, perceptions of robot anthropomorphism in the R-tutor group were significantly higher than those in the conventional teaching group. Behavior analysis results showed that the R-tutor group significantly outperformed the R-peer group and the conventional teaching group in terms of emotional empathy and behavioral empathy, while the interactions in the R-peer group and the conventional teaching group remained mainly at the level of cognitive empathy. The teacher’s teaching reflections indicated that robot tutors can stimulate preschoolers’ interest in learning, optimize their emotional experiences, and develop their empathy. The Situation-Cognition-Emotion-Behavior (SCEB) empathy learning model proposed in this study provides a theoretical framework for cultivating preschoolers’ empathy. Furthermore, the exploration of using HIRs with different roles to foster preschoolers’ empathy also offers valuable practical experience for preschool teachers in conducting empathy education.
In the process of collaborative learning, effective grouping is the key to improving learning efficiency. A well-structured collaborative learning group can markedly improve the learning efficiency of both individuals and the group as a whole. Nonetheless, current grouping methods for collaborative learning frequently fall short of a comprehensive assessment of students’ knowledge proficiency, thereby failing to guarantee that group members’ knowledge structures are mutually complementary. Therefore, a grouping model combining knowledge complementarity and dynamic diagnosis is proposed. The grouping model is divided into five layers. Firstly, the learning objective analysis layer deconstructs the teaching content into several core knowledge points. On this basis, the dynamic knowledge diagnosis layer uses the improved deep knowledge tracking model DKVMN-KT to dynamically diagnose students’ mastery of core knowledge points. Then, in the personalized grouping layer, K-means algorithm is used to cluster the personalized grouping results. Furthermore, teachers were involved in the manual fine-tuning of the grouping results. Finally, the model forms a closed-loop feedback link through a deep interactive optimization layer to promote the iterative optimization of the collaborative learning grouping model. The experimental results indicate that the collaborative learning grouping model is helpful to effectively group students at the level of knowledge structure, and can improve the efficiency of collaborative learning, In order to guarantee more equitable and diverse grouping outcomes, this methodology promotes positive interplay among students, enabling them to mutually learn from one another and thereby enhance their comprehension of diverse knowledge points effectively.
Recent advancements in Artificial Intelligence-powered robots have introduced a variety of educational technologies aimed at accommodating standard varieties of languages. However, the adaptiveness of the knowledge bases of these robots in establishing adaptive language learning at universities remains underexplored. This quasi-experimental study involved 1041 students from 13 academic disciplines with diverse first languages. Utilizing a parallel design, while students in the control group learned medical English skills through non-adaptive AI-powered robots, students in the experimental group learned the same skills through adaptive AI-powered robots. We assessed the effectiveness of learning medical English through Artificial Intelligence-powered robots with(out) adaptive knowledge bases, employing Objective Structured Video Exams for formative assessment. Additionally, the participants shared their perceptions of Artificial Intelligence-powered robots in medical English education through semi-structured interviews. Quantitative findings indicated that students using adaptive Artificial Intelligence-powered robots showed a greater sociolinguistic competence in medical English compared to those learning with non-adaptive robots. The qualitative findings suggested that the students in disciplines requiring direct interaction with patients and healthcare staff expressed a more positive perception of learning medical English through Artificial Intelligence-powered robots. While both groups acknowledged the value of the course, those engaging with adaptive systems demonstrated a great appreciation for the learning process and outcomes. Based on these results, we offer recommendations for researchers on effectively integrating Artificial Intelligence-powered robots into university language education, highlighting the potential for enhanced learning experiences through adaptive technologies.
With the increasing adoption of educational robots (ERs) to support creativity development, a comprehensive understanding of their research landscape, empirical effectiveness, and instructional conditions becomes necessary. This study addresses this need by proposing a three-stage analytical framework that involves topic modeling, meta-analysis, and rule mining. First, a BERTopic modeling approach was applied to identify four major research topics, including STEAM motivation and gamified project-based learning, creative coding and narrative play in early childhood, pedagogies for creative and computational thinking, and creative engineering and physical computing, with a growing emphasis on narrative and constructionist approaches, especially in early childhood education. Second, a Bayesian meta-analysis was utilized to identify a moderate overall effect size (Mean = 0.600), with subgroup analyses demonstrating that effect sizes varied significantly across experimental design factors such as subject domain, age group, learning duration, setting, instructional strategy, and measurement issue, where studies focusing on programming education, preschool and high school learners, and guided learning yielded the strongest effects. Finally, an associate rule mining approach was used to identify consistent patterns linking skill modeling, robot-as-tool roles, and mixed instructional strategies with positive creativity outcomes. Findings of this study contribute to both research and practice by enhancing researchers’ and instructors’ understanding of ER research trends and practical effectiveness in promoting creativity, with actionable instructional design recommendations being provided to optimize ER-based creativity interventions.
This special issue examines the landscape of robotics-facilitated teaching and learning through the theoretical lenses of embodied cognition and multimodal perspectives. The ten selected papers evidently report how both physical and virtual robots foster meaningful learning through social presence, multisensory interaction, pedagogical integration, and adaptive feedback. Collectively, these studies underscore the efficacy of synergizing robotic technologies with complementary pedagogical strands, ranging from scaffolding strategies (such as digital storytelling, concept mapping, and gamification) to collaborative and embodied processes (including socially shared regulation of learning and multimodal interaction), while also incorporating dialectic and inquiry-driven methods (such as Socratic dialogue and productive failure). Furthermore, the contributing insights reveal that well-orchestrated human-robot interaction yields multifaceted benefits across diverse age groups and subject domains, concurrently enhancing students' cognitive faculties (e.g., computational thinking, critical problem-solving, and creativity) and non-cognitive growth (e.g., empathy, socio-emotional learning, and professional identity).
This study introduces RobotPython, a software-based system with virtual robot assistants designed to support self-regulated learning (SRL)-based Python learning activities for programming novices in an undergraduate non-computer science context. RobotPython includes two virtual robot assistants: the Coding Scaffolding Guidance Assistant and the Flowchart Reflection Assistant, which guide students to learn Python in the performance and self-reflection phases of SRL respectively. Both groups received SRL-based instruction; the key difference was that the experimental group was supported by RobotPython, whereas the control group was supported by teaching assistants (TAs). The 22 students in the experimental group, who used RobotPython as an aid to instruction, achieved significantly higher scores in programming concepts and coding than the 28 students in the control group. The impact of RobotPython on SRL abilities was nuanced: the effectiveness of RobotPython for goal-setting, help-seeking, and self-evaluation was moderated by students’ baseline abilities (e.g., task strategies and time management), rather than showing a uniform improvement across all subscales. These results suggest that RobotPython-supported SRL-based instruction is associated with higher programming concept understanding and coding performance, while its impact on SRL abilities is conditional upon students’ existing characteristics. They also reveal that programming concept understanding mediates the relationship between SRL abilities and coding performance.
The use of chatbots has attracted much attention because it can create interactive and immersive learning environments for EFL (English as a Foreign Language) learners. While learners benefit from interactions with chatbots, they indicate that chatbots still have limitations in understanding their intent and expressions to maintain natural conversations. To provide learners better experiences, our team created the CoolE Bot based on generative artificial intelligence (GenAI), which helps improve the bot’s communication capabilities. This study investigated the effects of CoolE Bot on EFL learners’ listening and speaking skills, as well as its perceived usefulness across two proficiency levels. It adopted a pre- and post-test design involving 42 students. They took listening and speaking tests, engaged with CoolE Bot for seven weeks, and completed a survey. Statistical analysis indicated that, while both high- and low-proficiency groups showed significant improvements in speaking skill, the improvements in listening skill were less pronounced and did not reach statistical significance. In terms of perceived usefulness, the low-proficiency group found CoolE Bot more beneficial than their high-proficiency peers across several aspects, including different language skills development and increased opportunities for speaking practice. These results suggest that CoolE Bot is a valuable tool for enhancing speaking skills and supporting listening development.
This study developed and validated a scale for assessing online learning readiness among students aged 10 to 12 years in grades 4 to 6. The scale is called the Online Learning Readiness Scale for Elementary School (OLRS-E). To establish the reliability and validity of the scale, expert insights were gathered using the fuzzy Delphi method, and exploratory and confirmatory factor analyses were conducted on separate samples of 152 and 423 students, respectively. The OLRS-E comprises four key dimensions. Self-directed learning measures students’ motivation and ability to regulate their own learning behaviors. Communication self-efficacy assesses students’ confidence in expressing themselves and interacting in online environments. Family support captures the practical and emotional assistance students receive from caregivers during online learning. Technical support reflects students’ perceived competence in using digital tools and resolving basic technological issues. Among these, technical support yielded the highest score, indicating a high degree of technological readiness among Taiwanese elementary school students, followed by family support and communication self-efficacy. By contrast, self-directed learning yielded the lowest score, indicating a lower degree of self-directed learning readiness among these students. Overall, these findings of student readiness have major implications for parents and teachers regarding online learning efficacy.
Previous studies have shown that Social Shared Regulated Learning (SSRL) can facilitate knowledge construction and collaborative learning in STEM courses. Robots, as learning partners, play a crucial role in supporting SSRL processes and promoting collaboration. In addition, the game elements in gamification can effectively stimulate students’ learning motivation, promote changes in student behavior, and improve learning outcomes. Based on this, this study proposed a gamified robot-assisted SSRL approach and evaluated its effectiveness in terms of improving students’ academic performance, learning motivation and higher-order thinking through a quasi-experiment. The experimental group (n = 30) used a gamified robot-assisted SSRL approach (GR-SSRL), while the control group (n = 27) used a conventional robot-assisted SSRL approach (R-SSRL). The research results showed that the GR-SSRL approach can effectively improve students’ academic performance and external learning motivation, and can significantly improve the creative thinking ability of students with low collaboration ability. This study shows that the integration of certain learning tools can effectively support SSRL, and indicate that the combination of a variety of technical means can produce synergistic effects to improve learning outcomes.
As artificial intelligence (AI)-assisted writing tools are increasingly adopted, their impact on students’ critical thinking (CT) skills has garnered attention. However, behavioral and cognitive differences among different CT students using AI-assisted writing tools remain underexplored. This study developed multimodal AI-assisted writing agent (AI-AWA) based on Socratic dialogue principles and grounded in sociocultural theory. It analyzed the student-AI conversation transcripts and essays of 45 elementary school students using repeated measures analysis of covariance and Epistemic Network Analysis. It is found that High-CT students outperformed Low-CT students in terms of academic achievement. High-CT students demonstrated more pronounced improvements in grammar, whereas Low-CT students exhibited greater gains in content and organization. In terms of behavioral patterns, the High-CT group frequently posed challenging questions and independently constructed their own outlines, while the Low-CT students exhibited an increase in reasoning behaviors and a reduction in AI dependency. With respect to cognitive patterns, the High-CT students maintained advanced cognitive engagement, while the Low-CT students showed gradual improvement. These findings highlight how students with different CT levels engage differently with the AI agents, suggesting its potential for supporting diverse writing and thinking strategies. This study suggests that heuristic AI tools need to be designed differentially, using open scaffolding for High-CT students to stimulate their deep thinking, and more structured cognitive scaffolding for Low-CT students to balance tool support with CT autonomy. Such strategies allow for the dynamic maintenance and cultivation of critical thinking alongside writing support.
Educational Robotics (ER) courses are widely acknowledged as an effective method for fostering K-12 students’ Computational Thinking (CT). Previous teaching practices predominantly relied on unimodal visual instructional imitation and procedural task-based approaches. These approaches disconnect abstract computational symbols from embodied experiences and limit the in-depth development of students’ CT. To address this challenge, the present study proposes and systematically validates the Multimodal Embodied Interaction (MEI) model through an integrated perspective that combines embodied cognition with multimodal theory. A two-stage experimental study was conducted in a primary school in Southwestern China employing Design-Based Research (DBR) methodology. Results indicated that, compared to conventional teaching models, the MEI model significantly improved students’ CT skills, with especially notable impacts on higher-order thinking aspects such as abstraction, algorithms, and evaluation. The study also revealed that the model exerts a significant “protective effect” in maintaining students’ attitudes toward CT. This research provides a theoretically grounded and empirically validated framework for deepening students’ CT cultivation in ER courses.
The study aimed to examine the moderated mediation model for the influencing factors of adult assessors’ behavioral performance in online assessment (OA). The participants included 200 graduate students from a university seminar course utilizing an OA system. The results indicate that OA continuance intention fully mediates the relationships between the three antecedents (i.e., OA attitude, OA subjective norm, and perceived OA behavioral control) and OA behavioral performance. OA experience positively moderates the relationship between OA continuance intention and OA behavioral performance, and further moderates the mediating effect of OA continuance intention between OA attitude and OA behavioral performance, indicating a moderated mediation effect. This study expands the perspective and applicability of the theory of planned behavior (TPB) and provides practical implications for online peer assessment.
The rapid advancement of generative AI (GenAI) technologies, such as ChatGPT, has significantly transformed the educational landscape. Previous research highlights the impact of teachers’ beliefs on their professional development and instructional methods. There is a growing need to align teachers’ beliefs with the latest technological advancements. This study investigates the relationships between teachers’ epistemic beliefs and their self-regulated learning under GenAI contexts. With a sample of 687 educators, this research employs exploratory and confirmatory factor analyses to validate the GenAI Epistemic Beliefs Questionnaire and the GenAI Self-regulated Learning Questionnaire. Structural equation modeling reveals that specific dimensions of GenAI epistemic beliefs, particularly the justification for GenAI-generated content and the structure of GenAI-generated content, significantly influence the planning phase of self-regulated learning. Teachers with sophisticated beliefs in justifying GenAI-generated content demonstrate more effective planning strategies. In contrast, those with structured beliefs in GenAI-generated content exhibit a more rigid approach to planning. However, these epistemic beliefs do not directly impact the adaptation phase of self-regulated learning; instead, planning serves as a full mediator, emphasizing the critical role of initial self-regulated learning planning in fostering adaptability. This study underscores the importance of understanding how educators’ beliefs about GenAI influence their self-regulated learning.
The development of social and emotional learning (SEL) competence is vital for pre-service teachers’ professional identity. Although generative AI (GenAI)-supported digital storytelling can promote this development, it may also induce cognitive load and technostress if not properly guided. Therefore, this study proposed a self-explanation-integrated CIDI (Clarify, Ideate, Develop, Implement) approach to facilitate SEL internalization and ease GenAI-related stress. It explored participants’ perceptions of human-AI interaction through metaphor analysis and interviews. A total of 177 pre-service teachers were recruited for the study and assigned to either the S-CIDI (self-explanation-integrated CIDI) group (n = 90) or the C-CIDI (conventional CIDI) group (n = 87). Results showed that the S-CIDI model significantly improved learning performance and enhanced professional identity. Although social-emotional competence (SEC) improved in both groups, stress mindset increased across all conditions. Metaphor analysis and interviews revealed that the S-CIDI group exhibited weaker critical reflection on the AI-generated content, suggesting that caution is needed against the potential suppression of autonomous thinking by highly structured support. In metaphorical representations, both groups commonly described AI agents as “Assistants.” The S-CIDI group preferred authoritative metaphors such as “Tutor,” whereas the C-CIDI group tended to use metaphors like “Partner,” emphasizing human-AI collaboration and co-creation concepts. The interviews further revealed that although both models promoted the development of social-emotional competence and professional identity, they showed differences in technology perception and agency construction, with the S-CIDI model potentially fostering over-reliance and diminishing critical autonomy.
Serious games are recognized for promoting knowledge and skill transfer through active learning aligned with constructivist principles. However, the availability of guidelines on how specific game design elements align with these principles are limited. This study conducted a systematic literature review using Web of Science and ScienceDirect, analyzing 39 selected studies to profile trends in constructivist-serious games and identifying aligned GDEs. Findings showed that most studies focused on STEM subjects, targeted university students, used quantitative methods, and favored computer-based platforms. The most commonly applied theories were situated learning and problem-based learning. The most frequently used and strongly supported GDEs were “imposed choices”, “sensations”, and “objectives”. This highlights the importance of learners’ autonomy and clear educational objectives within an authentic learning environment in designing a constructivist-serious game. Each element was evaluated against four constructivist factors: interaction, problem-solving, active learning, and discovery, resulting in a conceptual mapping of GDEs to constructivist learning. These findings provide practical guidance for integrating constructivist principles into serious game design.
Cancer incidence has been rising among younger populations, making it increasingly important to educate adolescents about the relationship between diet and chronic diseases. However, conventional teaching methods often fail to promote deep and self-directed learning. Robot-based digital storytelling offers an engaging approach, but research shows that without effective scaffolding, students may struggle to construct coherent, knowledge-rich narratives. This study proposes a time sequence-oriented concept mapping approach to support high school students in designing robot storytelling projects focused on cancer-preventing dietary knowledge. A quasi-experimental design was adopted, with 19 students in the experimental group using time sequence-oriented concept mapping for robot-based digital storytelling and 19 students in the control group following a conventional robot-based digital storytelling model. Results indicated that the time sequence-oriented concept mapping-supported group outperformed the control group in learning achievement, communication tendencies, collaborative learning, critical thinking, and group self-efficacy. Interview findings further highlighted the model’s benefits in supporting knowledge construction and teamwork. This study provides empirical support for integrating time sequence-oriented concept mapping into robotics-supported health education.
While Content and Language Integrated Learning (CLIL) and generative artificial intelligence (GenAI) have received increasing attention in bilingual education, limited research has compared how different instructional frameworks operate within GenAI platforms. This study examines the effectiveness of Inquiry-Based Learning (IBL) and Adaptive Learning (AL) implemented through ChatGPT in enhancing fifth-grade CLIL students’ English vocabulary acquisition, content understanding, and scientific literacy in Taiwan. A quasi-experimental design involved 69 students across two groups: Experimental Group 1 (EG1, n = 34) used an IBL approach, and Experimental Group 2 (EG2, n = 35) followed an AL framework. Multimodal assessments included pre- and post-tests, PowerPoint designs, and oral presentations to evaluate students’ scientific vocabulary acquisition, content understanding, and literacy. Results indicated that EG1 significantly outperformed EG2 across outcome measures. On written post-tests, EG1 students showed stronger gains in vocabulary and content knowledge. Their PowerPoint designs demonstrated greater conceptual depth, accurate use of terminology, and more purposeful integration of visuals. Similarly, EG1’s oral presentations featured clearer scientific explanations, more fluent vocabulary use, and stronger alignment between visual and verbal elements. Qualitative analysis of rater reflections further revealed that EG1 students synthesized and communicated scientific knowledge more effectively through multimodal strategies. These findings suggest that GenAI-supported IBL offers advantages over AL in fostering bilingual learners’ cognitive engagement and scientific communication. This study contributes to the emerging literature on AI-enhanced bilingual instruction by showing how structured inquiry, supported by generative technologies, can advance the dual goals of content mastery and language development in CLIL science education.
