Aesthetic perception, the cognitive process through which individuals interpret and evaluate the expressive and emotional qualities of visual art, is fundamental to students' creative and emotional development. Recent progress in artificial intelligence has enabled computational models to assist in aesthetic analysis by identifying patterns in visual composition and affective expression. However, such models often struggle to recognize abstract or context-dependent aesthetic dimensions, and improving these aspects through comprehensive annotation remains costly and time-consuming. This study presents a human-AI teaming framework designed to identify the aesthetic perception dimensions that AI models find most difficult to interpret and to allocate these dimensions to human experts for annotation. The framework employs a multi-agent reinforcement learning (MARL) mechanism, where each agent is assigned to a specific aesthetic dimension and learns a policy for determining whether expert annotation is required. Two complementary state representation strategies are introduced: a statistical representation that captures the model's predictive distribution across dimensions, and a graph-based attention module that models interdependencies among aesthetic attributes. A reward mechanism further guides agents to balance the improvement of model perception with the minimization of human annotation effort. Experiments conducted on two real-world datasets demonstrate that the proposed framework effectively identifies the challenging dimensions for AI models and strategically delegates them for human evaluation. This targeted collaboration significantly enhances annotation efficiency and model interpretability, providing a scalable approach for improving human-AI synergy in aesthetic perception analysis.

Academic self-esteem (ASE) plays a significant role in children's learning outcomes by influencing their engagement and performance in subjects, such as math. Although existing studies emphasize the importance of self-esteem in academic achievement, the potential of interactive technologies, particularly virtual robots, in enhancing ASE remains underexplored. This study investigates the impact of a virtual robot on children's ASE, math performance, concentration, and engagement in an e-learning environment. The study involved an experimental group (<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</i> = 12) interacting with a virtual robot integrated into the math e-learning platform, and a control group (<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</i> = 12) working in a traditional e-learning setting without robot interaction. The results demonstrated that the experimental group, which interacted with the virtual robot, exhibited significant improvements in math performance, concentration, and engagement across the three experimental sessions (Sessions 1, 2, and 3) compared with the control group, as indicated by both quantitative measures and qualitative feedback from participants. ASE, as well as the quantity and quality of friendships, was assessed pre-and post-test, with findings indicating greater improvements in the experimental group after the intervention. The correlation between improved math performance and higher ASE was moderate to strong. These findings underscore the potential of virtual robots as tools that positively influence the ASE and learning outcomes of children, and highlight their value for future educational settings where such technology could address achievement gaps in mathematics learning through improved self-perception.

Sign languages are used by millions of Deaf and hard-of-hearing individuals worldwide, yet learning them can be challenging due to limited access to qualified instructors and real-time feedback. Sign language learning is crucial for inclusion and accessible communication, yet existing e-learning solutions often fail to provide detailed or explainable feedback. To address this gap, we present a novel computer-assisted system that leverages computer vision and machine learning to teach sign languages. In our approach, students follow pre-recorded videos of an instructor performing signs, and the system captures the students' own sign production through a conventional webcam. By analyzing both static handshapes and dynamic trajectories, our method provides detailed and explainable feedback at the level of keyframes, pointing out specific misalignments in hand, wrist, and motion path. This differs from many existing tools that offer only binary correct/incorrect evaluations without guidance for improvement. We implemented the system as a prototype with a full curriculum and conducted a pilot usability evaluation with 43 participants, ranging from beginners to skilled signers. Results indicate that users found the platform intuitive, the feedback mechanism informative, and the overall experience motivating for continued practice. The pilot assessed user perceptions of usability and feedback quality; learning outcomes were not measured in this initial evaluation. Notably, our design is language-agnostic, enabling broad application across different sign language communities. By offering instant, fine-grained feedback, our system aims to improve self-guided learning of sign languages and promote greater accessibility, ultimately bridging communication gaps and fostering inclusion for Deaf communities.

Training healthcare professionals in intravenous (IV) needle insertion is a critical component of medical education, traditionally relying on manikin-based simulations and real-life practice. However, the advent of Haptic Virtual Reality (HVR) technologies offers a transformative approach to this training, enhancing safety and potential efficiency. This study explores the development of an IV needle insertion simulator using two different haptic devices integrated into a VR system on the Unity platform and assesses its impact on learning through a 3-week experiment. The simulator is designed to create a realistic and immersive training environment by incorporating detailed anatomical models, physics-based hand interactions, and realtime haptic feedback. The virtual environment replicates a clinical setting, featuring a patient arm model and an IV needle. The haptic feedback is programmed to offer realistic feelings of needle insertion and hand grasping, improving the user's accuracy. Learning impact and usability testing with 41 students indicate a promising improvement in skill acquisition and confidence. Specifically, participants showed a 55% increase in success rates and a significant boost in confidence. This highfidelity HVR simulator represents a significant step forward in medical training technologies, offering a scalable and repeatable training tool adaptable to various educational needs and skill levels.

According to the ICAP framework, engaging in <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">constructive</i> cognitive modes yields better learning outcomes than <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">active</i> modes. Also, prior studies on educational chatbots suggest that enhancing chatbot humanness can improve learning. However, these two ideas have not been fully explored together, especially within the context of text-based, disembodied chatbots. This study investigates the impact of the cognitive engagement modes (<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">constructive</i> vs <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">active</i>) and chatbot humanness (humanized vs non-humanized) on learning outcomes and five dimensions of learning motivation. We conducted a two-by-two factorial user experiment with 55 chatbot users. Data was analyzed through a mixed-method approach to examine the main and interaction effects of the two independent variables. Regarding learning outcomes, our data showed that learners who interacted with <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">constructive</i> chatbots showed higher learning outcomes than those who interacted with <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">active</i> chatbots. In addition, learners who interacted with non-humanized chatbots reported higher learning outcomes than those who interacted with humanized chatbots. Lastly, we observed a significant interaction effect between the two independent variables on tension-pressure and perceived competence, which are two dimensions of learning motivation. Our study extended the applicability of the ICAP framework to the domain of chatbot-based learning, challenged the assumption that the humanness of chatbots can lead to improved learning outcomes, and underscored the importance of exploring both the cognitive engagement modes and the humanness of chatbots when designing chatbots to enhance users' learning motivation.

Immersive extended reality (XR) technologies, including augmented reality (AR), virtual reality, and mixed reality, are transforming the landscape of education and training through experiences that promote skill acquisition and enhance memory retention. These technologies have notably improved decision making and situational awareness in public safety training. Despite the promise of these advancements, XR adoption for emergency response has been slow. This hesitancy can be partially attributed to a lack of guidance for integrating these novel technologies into existing curricula. This work aims to guide instructional designers, curriculum developers, and technologists in seamlessly integrating immersive technologies into public safety training curricula. This work provides a comprehensive account of our collaboration with instructional designers, public safety personnel, and subject matter experts in developing an AR-based training curriculum for the Sort, Assess, Life-saving Interventions, Treatment/Transport triage technique used in mass casualty incidents (MCIs). In addition, we introduce a systematic framework for public safety curriculum development based on the Analyze, Design, Develop, Implement, Evaluate instructional design model. Leveraging a human-centered design approach, we first analyze the necessity for immersive learning in public safety. Next, we identify the obstacles in developing XR training experiences and outline our construct of a training prototype through iterative evaluations based on stakeholder feedback. Finally, we share qualitative insights through iterative evaluations with firefighters and emergency medical technicians performing MCI triage tasks in AR, supplemented by survey questionnaires and semistructured interviews. Our goal is to provide a blueprint for a successful integration of immersive technologies into public safety training curricula.

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.

The capabilities of large language models (LLMs) in language comprehension, conversational interaction, and content generation have led to their widespread adoption across various educational stages and contexts. Given the fundamental role of education, concerns are rising about whether LLMs can serve as competent teachers. To address the challenge of comprehensively evaluating the competencies of LLMs as teachers, a systematic quantitative evaluation based on the competency model has emerged as a valuable approach. Our study, grounded in the teacher competency model and drawing from 14 existing scales, constructed an evaluation framework called TeacherComp. Based on TeacherComp, we evaluated six LLMs from OpenAI across four dimensions: knowledge, skills, values, and traits. Through comparisons between LLMs' responses and human norms, we found: 1) With each successive update, LLMs have shown overall improvements in knowledge, while their skills dimension scores have increasingly aligned with human norms. 2) There are both commonalities and differences in the performance of various LLMs regarding values and traits. For instance, while they all tend to exhibit more negative traits than humans, their morals can vary. 3) LLMs with reduced security, constructed using jailbreak techniques, exhibit values and traits more closely aligned with human norms. Building on these findings, we provided interpretations and suggestions for the application of LLMs in various educational contexts. Overall, this study helps teachers and students use LLMs in appropriate contexts and provides developers with guidance for future iterations, thereby advancing the role of LLMs in empowering education.

Predicting student engagement can provide timely feedback and help teachers make adjustments to their practices to meet student needs and improve their learning experience. This article proposes a four-step approach using a sequential ensemble model for engagement prediction, discusses the contribution of different features to the model and the influence of video segmentation in the prediction, reports on two in-the-wild datasets-The Emotion Recognition in the Wild Engagement Prediction (EmotiW-EP) dataset published in 2018 as part of a student engagement task and the Dataset for Affective States in E-Environments (DAiSEE), a general purpose dataset also used in the educational context but not limited to it, and, finally, presents a comprehensive and thorough critical analysis, highlighting crucial factors to consider when using AI/computer vision models in educational datasets for learning purposes. Experiments show that our proposed approach outperforms state-of-the-art approaches by obtaining a mean square error of 0.0386 on the DAiSEE dataset and 0.0610 on the EmotiW-EP dataset. We conclude this article with a critical analysis of the reliability of such predictions in learning environments and propose future directions for the effective use of AI/computer vision models in education.

Teaching programming is a topic that has generated a high level of interest among researchers in recent decades. In particular, multiple approaches to teaching visual programming have been explored, from the use of tools such as Scratch, robots, unplugged programming or activities for the development of computational thinking. Despite the wide range of resources used, students generally tend to perform poorly academically and perceive learning visual programming as a complex and demotivating task. <p xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">In this article, the TutoApp system is proposed together with a new methodology based on “Learning by Teaching”, where students create tutorials in their mobile devices to explain programming concepts to their peers. The hypothesis of this paper is that the use of this tool improves learning outcomes and the level of student satisfaction. An experiment with a pre-post-test design has been carried out with 57 university students in an introductory programming course, 30 belonging to a control group (did not use TutoApp) and 27 belonging to the experimental group (used TutoApp). The findings indicate that the creation of tutorials with TutoApp significantly improved students' academic performance over those who did not use it, specifically in learning the loops and conditional control structures. However, it was observed that anxiety increased in all students while learning visual programming. <p xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">The results of this study open the door to the validation of the use of systems and methodologies for creating tutorials for teaching visual programming to university students.