Immersive virtual reality (iVR) provides authentic contexts for second-language speaking, yet learners often encounter high-risk moments (HRM) such as prolonged pauses, repeated help-seeking, or error–help loops disrupting self-regulation. Although scaffolding is widely used to address these difficulties, little is known about when support should be triggered in iVR. This study examined whether sequential models can anticipate HRM in a GenAI-supported iVR speaking task and which behavioural signals best explain these predictions. Event logs from 80 undergraduates interacting with a digital human were coded into 14 self-regulated learning (SRL) categories. HRM were operationalised through repeated help-seeking, error–help cycles, and inactivity beyond 10 seconds. Predictive features were derived from sliding windows of prior SRL events and tested with logistic regression, Markov chains, long short-term memory (LSTM) networks, and temporal convolutional networks (TCN). Results showed that the TCN achieved the strongest performance (PR-AUC = 0.60; ROC-AUC = 0.76), successfully flagging most HRM. SHAP and TimeSHAP analyses revealed interpretable precursors including help bursts, entropy collapse, and long pauses. Findings illustrate a decision-aligned learning analytics pipeline that links prediction with interpretability and scaffolding design. By translating predictive signals into operational rules, the study highlights how learning analytics can support just-in-time interventions in iVR language learning.

The rise of the metaverse as an educational platform has introduced new opportunities and challenges in understanding students' cognitive processes. Traditional learning analytics methods often fail to fully capture the dynamic patterns of cognitive presence and metacognitive activities in immersive virtual environments, particularly when students interact with an artificial intelligence (AI)-powered digital human. To address this gap, this study aims to develop a multimodal learning analytics (MMLA) approach to analyse cognitive presence and metacognition in Learningverse - a metaverse platform. To address the limitations of traditional methods, this research integrated eye-tracking data with dialogue text from the AI-powered digital human to provide a more comprehensive understanding of these patterns. A pilot study involving undergraduate students was conducted to collect data, revealing specific patterns of cognitive presence and metacognition. The findings suggest that the MMLA approach could offer deeper insights into students' learning behaviours, providing valuable implications for the design of educational tools and the development of more effective learning strategies in virtual environments.

The growing use of virtual learning environments (VLEs) in primary education offers new opportunities for enhancing student learning. However, understanding and analysing student behaviour in these environments is still challenging, especially in collaborative science education. This research aims to develop and evaluate a multimodal learning analytics (MMLA) approach tailored for primary science education. The study will focus on four key questions: understanding the current state of learning analytics (LA) in VLEs, identifying which data types in MMLA most effectively contribute to insights into student learning behaviours, creating an MMLA approach to support collaborative problem solving (CPS) and metacognitive strategies in VLEs, and improving the visualization of MMLA results for educators and students. To achieve these goals, the research will use a series of studies that collect and analyse multimodal data, including eye-tracking, behaviour logs, and dialogue text. Machine learning and deep learning techniques will be applied to identify critical data types, and these insights will inform the creation of an MMLA approach specifically designed to support CPS and metacognitive strategies. A user-centred design will guide the creation of a visualisation dashboard. This research is expected to contribute to theory by expanding the application of the MMLA approach, critically reflecting on the potential to innovate CPS and metacognitive strategies within VLEs and improving data analysis and dashboard design, and to practice by enhancing tools for primary science education.

In the current study, we explored the learning outcomes of five learners with different levels of metacognitive awareness in inquiry-based learning activities within the metaverse, and used eye-tracking technology to reveal the changes in eye movements of different learners during the learning process. The results showed that learners with high metacognitive awareness levels dealt with problems more comprehensively in general during problem-solving than learners with low metacognitive awareness levels. In addition, eye movement data showed that students with high metacognitive awareness levels tended to identify the inquiry question first and then carry out specific activities during their interactions; at the same time, they also had a more reasonable sequence of interactions with the learning materials compared to students with low metacognitive awareness levels. The study has practical implications for the future design of inquiry-based learning activities in the metaverse.

Traditional digital learning environments faced challenges in obtaining comprehensive user interaction data, often yielding fragmented insights without a cohesive visual representation. The emergence of metaverse platforms has enriched this landscape, enabling detailed user activity representation with multimodal data through avatars. However, how to understand the multimodal data related to teaching, social and cognitive presences underpinned by the 'Community of Inquiry' theoretical framework in the metaverse is a big challenge for educators. This study introduces a holistic visualisation solution to bridge this gap, ensuring a better understanding of avatars’ behaviours in an educational metaverse platform – Learningverse developed by our research team. The solution captures a range of multimodal data in Learningverse, such as avatar location, behaviours, emotions, and conversation. Key visualisation elements include heatmaps, points, and arrows, each with distinct informational value. In the future, integrating the solution with multimodal learning analytics is our next step work to understand teaching, social and cognitive presences.

Despite that Augmented Reality (AR) has been integrated into English as a Foreign Language (EFL) instruction, few studies have been conducted, focusing on AR -supported vocabulary learning both inside and outside the classroom at a primary level. Against this background, this quasi-experimental study aimed to examine the impact of an AR app – VocabGo on primary students’ vocabulary learning engagement under the ‘pedagogical framework of AR-supported vocabulary acquisition in a seamless learning environment’ in the context of a private school in Shenzhen, China. Mixed research methods were adopted. The study lasted 22 weeks. Seventy-two Grade 4 students from a private school in Shenzhen were randomly divided into three groups (N = 24 per group). Group 1 used VocabGo both in-class and out-of-class, Group 2 used VocabGo in-class only, and Group 3 used VocabGo out-of-class only. Data collection involved student pre- and post-engagement questionnaires, student focus group interviews, and log data. The research findings show that students in Group 1 outperformed significantly those in Group 2 and Group 3 in learning engagement in cognitive, behavioural, emotional and agentic dimensions. This indicates that AR-supported vocabulary acquisition in a seamless learning environment is conducive to students’ vocabulary learning engagement in all dimensions. This study contributes to the literature in substantiating the effectiveness of AR-supported vocabulary learning in a seamless learning environment for young learners in an EFL context, particularly in Mainland China.

In this paper, we proposed a productive failure instructional design (Kapur, 2008) to enhance Hong Kong secondary students’ English grammar learning and collaborative problem-solving skills in a MCSCL environment. It is believed that the pedagogical implications of this study will make contributions to the technology-enhanced language learning field. In addition, insight gained from this study will hopefully extend to investigating the impact of productive failure pedagogical design on listening, speaking, writing, and reading skills in the L2 acquisition.

In this poster, we reviewed 38 articles on learning analytics research, focusing on the data visualization interface designs. After examining the original ideas their interface design, a new visualization strategy was proposed to categorize and characterize them premised on their principles and approaches in four types, namely, (1) directly-presented, (2) outcome-oriented, (3) process-oriented, and (4) theory-oriented. Then, how these types of the visualization strategy could help facilitate learning analytics design and make data interpretable by users was presented.

This study aims to investigate (1) assessment measures utilized in Computer-Supported Collaborative Learning (mCSCL) research; (2) whether these assessment measures have examined the effectiveness of mCSCL that the studies intend to assess; and (3) when the assessment measures are conducted in mCSCL research in an attempt to bring to light potential methods that are conducive to examining effectiveness of mCSCL practices and sustain the practices, and identify methodological issuesin mCSCL to be addressed in future research. The research findings show a variety of methodological issues that need to be addressed. Discussions are made in comparison with the findings in CSCL research and other studies leveraged by mobile technologies. Potential directions to investigate the effectiveness of mCSCL practices are proposed.

This paper reports an on-going case study on the project of “Bring Your Own Device (BYOD) for seamless science inquiry” in a primary school in Hong Kong. The study aims at investigating how the students advanced their content knowledge in science inquiry in a seamless learning environment supported by their own mobile devices. The topic of inquiry was “ The Anatomy of Fish”. Data collection included pre- and post-domain tests, student artifacts, class observations and field notes. Content analysis and a trialogical approach were adopted in the data analysis to trace the students’ knowledge advancement. The work of one group of students was used as an example. The research findings show that the students advanced their understanding of the anatomy of fish well beyond what was available in the textbook.

This paper presents “Move, Idioms!”, a Mobile-Assisted Language Learning design that emphasizes learners‟ habit of mind and skills in making meaning with their daily encounters, and associating those with the knowledge learned at the formal classes. The students used smartphones on a 1:1, 24x7 basis to take photos in real-life contexts related to Chinese idioms, made sentences with the idioms, and posted them onto a wiki space for peer reviews. In this paper, we focus on investigating students‟ cognitive processes and patterns in artifact creation. Through two case studies, we gained better understanding of how (1) physical settings in the informal contexts, (2) parental support, and (3) technology facilitated students‟ meaning making in their daily life. We hope to contribute to the literature of mobile learning by exploring virtually limitless learning opportunities that informal contexts and parental involvement may offer to learners.