The proliferation of generative artificial intelligence (GAI) tools such as ChatGPT has catalyzed growing interest in their potential to support personalized learning (PL) in educational settings. While the promise of PL lies in tailoring instruction to individual learner needs, the integration of generative AI into such pedagogical models remains under-explored and insufficiently evaluated. This study presents a two-phase design-based research examining the affordances and limitations of ChatGPT as a tutor bot in facilitating PL among graduate students. Through interaction analysis, rubric-based performance assessment, and behavioral flow visualization, the study reveals that AI-driven tutoring can significantly enhance learner outcomes—especially when guided by structured prompts and pedagogically aligned designs. Findings underscore the dual importance of instructional design and real-time feedback quality in leveraging AI tools for effective, personalized learning experiences. The study contributes to the evolving discourse on human–AI partnerships in education and calls for deeper integration of learning science principles in the deployment of generative AI technologies.

Autistic learners encounter a range of communication challenges in socialemotional communication, non-verbal communication, and building and sustaining relationships. This conceptual paper explores the impact of technology, particularly extended reality (XR), in aiding individuals with autism spectrum disorder (ASD) in enhancing their communication skills. It begins by providing a concise summary of the three main communication obstacles faced by individuals with ASD, and then proceeds to examine various technologies used for ASD education. The primary objective of this paper is to explore the relationship between XR and ASD education through the introduction of a five-component model. It concludes by offering practical recommendations for integrating XR into communication training for individuals with ASD.

The integration of large language models (LLMs) into intelligent tutoring systems promises personalized learning, yet procedural domains such as primary mathematics remain challenging due to inconsistency and pedagogical opacity. LLM tutors may give unsolicited answers or show unstable scaffolding when students vary arithmetic, units, or rounding. This design science research develops and evaluates a rule-guided tutoring system for mathematical word problems in primary schools. We formalize a distinction between rule-guided scaffolding, in which tutoring is governed by a three-layer architecture (diagnosis → intent selection → constrained response generation), and ad-hoc scaffolding, where helpful moves are difficult to audit and replicate. The artifact leverages LLMs for natural language generation while constraining stochastic variability through a structured pedagogical framework informed by scaffolding theory. Evaluation followed a staged Design Science Research (DSR) approach: (1) persona-based simulated student–tutor dialogues to identify failure modes, and (2) a classroom pilot with 40 Grade 5 students to test ecological robustness. Results indicate that rule-guided scaffolding improves interactional consistency, reduces premature answer-giving and early closure, and sustains student cognitive engagement. The classroom pilot further revealed interactional complexities, fragmented inputs, and attentional fluctuations, not captured in simulation, highlighting the value of staged evaluation. This study contributes an auditable architecture and empirically grounded design principles for transparent, reliable LLM-based tutoring in authentic classrooms.

PurposeWorkplace interaction and collaboration can be enhanced by networked learning. The study intends to explore networked learning in the workplace (knowledge sharing and connection buildings) and gain insights into how workers develop connections through learning analytics social network analysis (SNA).Design/methodology/approachSNA was employed to explore how learning connections were established amongst healthcare workers in a large hospital in Singapore. We examined both the total network interactions (density, diameter, average shortest path length) and the levels of interactions between individuals (degree, betweenness, closeness centralities). A total of 99 responses were included in the final data analysis, and Python packages such as NetworkX were used to perform SNA.FindingsThe network as a whole is sparse, as indicated by the low-density score (0.4%). The findings of the study reveal that the bigger sub-networks had more than one worker who interacted with more than one co-worker and these tend to have more edges in them interlinking workers from different departments. We also found that workers from the departments with the larger populations in the sub-networks were more likely to have the highest degree, betweenness and closeness centrality values. This indicates that the larger sub-networks hold more value in terms of understanding how workers with higher centrality values are nurtured.Originality/valueThis paper sheds light on the learning process that occurs when workers engage in networked learning and provides empirical findings with Singapore as the context of the study.

Gender disparities in Computational Thinking (CT) are well-documented, yet few meta-analyses examine how males and females respond differently to educational interventions. This three-level meta-analysis synthesized 37 studies (62 effect sizes) for females and 35 studies (56 effect sizes) for males to explore gender-specific patterns in CT education programs. Results show significant overall effects for both females (g = 0.867) and males (g = 0.807), confirming the effectiveness of CT education across genders within the reviewed studies. However, moderator analyses reveal that males demonstrated consistent gains across educational levels, assessment tools, and learning outcomes, while females showed more variation. Females improved most in CT performance and attitudes, but showed limited gains in self-efficacy and self-perceptions. Pedagogical approaches such as unplugged activities, robotics, and project-based learning were effective for both genders. The findings also suggest a need for age-appropriate and culturally relevant pedagogies, particularly for girls in primary and secondary education. While this meta-analysis provides practical implications for future program development, it also highlights the need for broader research across diverse sociocultural and technological environments to better understand and address the complex dynamics of gender in CT education.

The integration of Artificial Intelligence (AI) in education has underscored the urgent need to equip educators with essential AI literacy and related competencies. This paper highlights the critical importance of advancing research on the development of teachers' AI literacy, particularly through targeted professional development (PD) programs. The study piloted a PD program involving 19 mid-career teachers in Singapore. Over six sessions, each lasting three hours, participants engaged in an intensive program designed to enhance their ability to integrate AI into educational practices. The program placed a strong emphasis on ethical considerations and the responsible use of AI. Mixed methods were employed. Data collection included pre-and post-intelligent Technological Pedagogical Content Knowledge (TPACK) surveys, teachers' perceptions on AI, in-class group discussions, and written assignments. Data analyses included content analysis and quantitative data analysis. The results showed a significant enhancement in i-TPACK, accompanied by a noteworthy shift in their overall perceptions of AI. The teachers not only acquired a good understanding of ethical frameworks but also demonstrated adept application in envisioning innovative AI in teaching, schools, and assessment. Importantly, they formulated tailored action plans for implementing AI in their respective schools. The study employed a novel analytical matrix based on the Aristotelian tripartite division of knowledge-episteme, techne, and phronesis-to compare action plans between teachers with different perceived i-TPACK levels, focusing on AI's application in teaching, schools and assessment. This study contributes valuable insights into teacher PD concerning AI in education and informs the implementation of AI in teaching practices.

Rapid advancements in Artificial Intelligence (AI) have prompted growing interest in leveraging AI for educational feedback. Yet, the centrality of the learner in this process is often overshadowed by technological excitement, and a broad understanding of AI-assisted feedback (AIFB) in education remains evolving. To address this gap, we conducted a systematic review of 129 peer-reviewed journal articles (2014–2023) based on widely used AI-related search terms to examine how AI, especially generative AI, supports feedback mechanisms and influences learner perceptions, actions, and outcomes. Our analysis identified a sharp rise in AIFB research after 2018, driven by modern large language models. We found that AI tools flexibly cater to multiple feedback foci (task, process, self-regulation, and self) and complexity levels (basic, intermediate, and elaborated). Our findings demonstrate that AIFB can effectively enhance targeted learning outcomes. By employing a transparent and field-aligned methodology, we synthesized recent advances and offers actionable insights for both research and practice. While the focus on widely recognized AI-related search terms ensures strong comparability and relevance, some specialized subfields (e.g., Automated Writing Evaluation), are less prominent in this synthesis. The study also highlights the ongoing need for clearer reporting of underlying AI algorithms. Building on these findings, we propose an original conceptual model that synthesizes current progress and offers a roadmap for future explorations. By illuminating the affordances and constraints of AIFB, we highlight the necessity for transparent methodological reporting and underscores the importance of integrating pedagogical and technological insights to promote meaningful, learner-centered feedback.

This paper reports the an AI-enabled project in space education, which seeks to enhance k-12 school students' engagement in underexplored STEM domains through an intelligent AI-driven platform. By integrating Design Thinking and the Predict - Observe - Explain (POE) pedagogical framework, the project develops two core components: (a) a Subject-Specific Image Explanation Tool (SS-IET), fine-tuned on space education datasets using open-source large language models, and (b) a Subject-Specific Chatbot-based Learning Tool (SS-CLT) that scaffolds inquiry learning via adaptive question prompts. Preliminary results demonstrate that domain-specific fine-tuning improves explanatory accuracy, reduces hallucination, and aligns language style with the developmental needs of K - 12 learners. The paper discusses the project rationale, system design, interim findings, and outlines future research directions to strengthen the role of AI in advancing STEM education in space science.

Purpose Networked learning facilitates collaboration and learning interactions. This study aims to explore networked learning in the workplace (knowledge sharing and connection building) and gain insights into the contextual factors (learner and environmental) of learning interactions. Design/methodology/approach Thematic analysis was conducted to explore how learning interactions were facilitated among healthcare workers in a large hospital in Singapore. This study examined both the learner and environmental factors of learning interactions. Interview and focus group discussions qualitative data were included in the analysis. Findings The findings of this study reveal that more emphasis can be placed on the environmental factors, and targeting these factors would provide a good foundation for networked learning in the workplace, whereas learner factors could be promoted to enhance additional learning interactions. This study also found that workers learn most frequently from experienced seniors, indicating the value of mentorship programmes in fostering high-quality learning interactions. Originality/value This paper sheds light on the comprehensive set of factors that encourage networked learning among healthcare professionals and provides empirical findings that might direct future studies in similar domains.

Purpose The review aims to synthesize previous studies to present an overview of the techniques commonly used in learning analytics, as well as identify possible knowledge gaps in the extant studies and provide insights on future directions for learning analytics techniques moving forward. Design/methodology/approach This paper provides a systematic review of learning analytics techniques. A total of 63 articles were included in the final review and 3 main themes emerged based on our research questions. These themes include (A) individual learning, (B) collaborative learning and (C) game-based learning. The first theme is related to the application of learning analytics techniques in the context of individual student learning, while the second and third themes focus on the application of learning analytics techniques in the context of collaborative learning and game-based learning research, respectively. The paper summarizes key findings, identifies possible gaps for future research and provides recommendations for future research. Findings The commonly used techniques include classification, content analysis, social network analysis and taxonomic mapping. Multimodal learning analytics, which uses data from multiple sources to understand learners’ behavior and experience, is also growing. The review of learning analytics research highlights several knowledge gaps, including methodological issues, adaptability of techniques, ethical, risk and privacy concerns and precise terminologies for methodological decisions. The choice of learning analytics techniques should be guided by research questions and data nature. Originality/value This work meets the originality requirement.

This study examines how a scaffolding tutor chatbot elicits higher-order thinking (HOT) in a graduate-level course. We coded 50 anonymized chatbot–student turns for tutor prompt type, and HOT type (Critique and Transfer), and depth (low/high for student's response). Prompt categories were derived inductively and include Social Greeting, Invite Experience Recall, Invite Problem Identification, Invite Critical Reasoning, Invite Contextual Application, Invite Solution Generation, and Invite Summary Synthesis. Results show that not all prompts equally lead to high-quality HOT: Invite Critical Reasoning (Type 4) and Invite Solution Generation (Type 6) reliably produced Critique_High and Transfer_High responses; Invite Experience Recall (Type 2) frequently produced rich, high-level transfer; Invite Summary Synthesis (Type 7) also elicited deep, comprehension responses. By contrast, Invite Contextual Application (Type 5) often remained at low-level. We suggest that dialogue design shapes HOT depth. Practical implications are proposed, such as keeping the seven types of prompts and adding more scaffolding to Invite Problem Identification and Invite Contextual Application. Future work should broaden the dataset and examine more HOT indicators.

Since the advent of chatbots enabled by Generative AI such as ChatGPT, their application in the domain of education has been linked to promises of personalizing learning (PL). Through a study of conversational interactions of graduate students with such chatbots, this paper provides an empirical study of how current ChatGPT technologies can enable PL. We introduce a framework of levels of PL affordances enabled by the current state of ChatGPT technology. The case study uses multi-methods to analyze graduate student interactions with a ChatGPT tutor bot in discussions about the topic of educational reforms. Tutoring conversations of 51 students with the tutor bot are collected, and then analyzed to probe whether and how ChatGPT creates an individualized growth area for students, tailoring interactions to individual needs and supporting students to have a deeper understanding of the topic. This study contributes to the limited body of research on the evidence-based application of Generative AI to PL.

In the current era where computational literacy holds significant relevance, a growing number of schools across the globe have placed emphasis on K-12 programming education. This field of education primarily comprises two distinct modalities—the block-based programming modality (BPM) and the text-based programming modality (TPM). Previous research may not have provided a complete understanding of the differences between these two modalities as it did not take into account both the learning process and learning outcomes. This study aimed to compare secondary students’ programming behaviors, computational thinking skills, and attitudes toward programming between the two modalities through a quasi-experimental design in a Chinese secondary school. The findings showed that (1) learners in TPM encountered more syntactical errors and spent more time between two clicks of debugging, while learners in BPM had more code-changing behaviors by adjusting programming blocks, made more attempts of debugging, and had more irrelevant behaviors; (2) learners in BPM achieved a higher level of computational thinking skills; (3) learners in both modalities experienced a slight decrease in confidence and enjoyment, while learners in BPM had higher interest levels in programming. (4) Code Changer, Minimal Debugger, Maximal Debugger, Distracted Coder and Average Coder were identified through students’ programming behavior in the two programming modalities, and differences in their CT skills and attitudinal data were revealed. Lastly, pedagogical implications based on the findings are also discussed.