Discussion has been widely used in courses, both online and otherwise, as it provides opportunities for students to construct knowledge through interaction with peers and instructors. Grouping students is a prominent strategy in the use of discussion. However, simply dividing students cannot guarantee active participation and high learning performance. There is therefore a need to pay attention to the structure and/or features of grouping, especially group size and group composition. The study described in this article focuses on the combined effects of group size and group composition in online discussion. It investigates whether students in small groups have different participation behaviors and learning performance compared to students in whole-class discussion. In addition, the influence of group composition is examined by comparing students’ participation and learning performance from high, medium, and low social-connected groups. Furthermore, this study also investigates how students’ perceived learning experience differs among these three differently-connected group compositions. The results indicate significantly different participation behaviors and learning performance between small-group and whole-class discussion. The effects of group composition are also shown in students’ learning behaviors, performance, and perceived experience. The results also reveal both advantages and disadvantages of different group types. The findings are expected to inform the design and implementation of grouping methods and extend our understanding of online discussion.

Self-regulated learning (SRL) is a sustainable development skill that involves learners actively monitoring and adjusting their learning processes, which is essential for lifelong learning. Learning feedback plays a crucial role in SRL by aiding in self-observation and self-judgment. In this context, large language models (LLMs), with their ability to use human language and continuously interact with learners, not only provide personalized feedback but also offer a data-driven approach to sustainable development in education. By leveraging real-time data, LLMs have the potential to deliver more effective and interactive feedback that enhances both individual learning experiences and scalable, long-term educational strategies. Therefore, this study utilized a quasi-experimental design to examine the effects of LLM-based feedback on learners’ SRL, aiming to explore how this data-driven application could support learners’ sustainable development in computer programming learning. The findings indicate that LLM-based feedback significantly improves learners’ SRL by providing tailored, interactive support that enhances motivation and metacognitive strategies. Additionally, learners receiving LLM-based feedback demonstrated better academic performance, suggesting that these models can effectively support learners’ sustainable development in computer programming learning. However, the study acknowledges limitations, including the short experimental period and the initial unfamiliarity with LLM tools, which may have influenced the results. Future research should focus on refining LLM integration, exploring the impact of different feedback types, and extending the application of these tools to other educational contexts.

The proliferation of massive open online courses (MOOCs) highlights the necessity of developing accurate and diagnostic evaluation methods to assess the courses’ quality and effectiveness. Hence, this study proposes a diagnostic MOOC evaluation (DME) method that combines the Analytic Hierarchy Process algorithm and learner review mining to integrate expert opinions, standardized rubrics, and learner feedback into the evaluation process. For this purpose, 30 MOOCs from the Coursera website were purposively selected and evaluated using the DME method and the results compared with expert evaluation and learner rating scores. The preliminary findings, in general, support the feasibility, accuracy, and diagnostic utility of the DME method and its suitability as a low-cost, sophisticated, and accurate method for MOOC evaluation. Finally, the study discusses several limitations and technical issues of the DME method that should be addressed in future research and practice.

China’s education system faces a loss of teachers and a decline in education quality, and the new fertility policy means that the demand for teachers will increase. Against this background, the motivations and perceptions of pre-service teachers (N= 2618) were investigated. The Factors Influencing Teaching Choice instrument was used, and its construct validity and reliability were assessed. The results indicate that social altruism motivation was the most influential, in contrast with the high ratings of perceived teaching ability and intrinsic value motivations in Western studies. Prior teaching and learning experience came next, highlighting the prominence of the practices of teacher education programmes. Female pre-service teachers reported significantly higher motivations from intrinsic career value and shaping the future of children/adolescents and fewer fallback-career motivations. With increasing college year, the study found significantly higher values for perception of high task demand. Social utility values and expert career were scored higher by rural pre-service teachers than by urban ones. It is argued that policy makers, politicians and others who care about improving students’ intrinsic career value should enable pre-service teachers to undergo a longer internship time and targeted learning guidance.

Problem-based learning (PBL) has been used in different domains, and there is overwhelming evidence of its value. As an emerging field with excellent prospects, learning analytics (LA)—especially multimodal learning analytics (MMLA)—has increasingly attracted the attention of researchers in PBL. However, current research on the integration of LA with PBL has not related LA results with specific PBL steps or paid enough attention to the interaction in peer learning, especially for text data generated from peer interaction. This study employed MMLA based on machine learning (ML) to quantify the process engagement of peer learning, identify log behaviors, self-regulation, and other factors, and then predict online PBL performance. Participants were 104 fourth-year students in an online course on social work and problem-solving. The MMLA model contained multimodal data from online discussions, log files, reports, and questionnaires. ML classification models were built to classify text data in online discussions. The results showed that self-regulation, messages post, message words, and peer learning engagement in representation, solution, and evaluation were predictive of online PBL performance. Hierarchical linear regression analyses indicated stronger predictive validity of the process indicators on online PBL performance than other indicators. This study addressed the scarcity of students’ process data and the inefficiency of analyzing text data, as well as providing information on targeted learning strategies to scaffold students in online PBL.