Computer-supported, robot-assisted collaborative programming (CSRACP) is a computer-supported collaborative learning (CSCL) mode used in programming education. CSRACP allows students to work in groups with the support of robots to achieve programming functions and complete complex tasks. This research applied multimodal learning analytics (MMLA) to examine the effects of three pedagogical scaffoldings (i.e., conceptual, task-oriented, and meta-cognitive scaffolding) on student pairs’ CSRACP activities in higher education. The results revealed that students under conceptual scaffolding had the low-level cognitive engagement, task-oriented regulation, and low-level socio-emotional expression, with the lowest programming task score; students under task-oriented scaffolding had the operation-driven cognitive engagement, observation-oriented regulation, and low-level socio-emotional expression, with the highest programming task score; and students under meta-cognitive scaffolding had the communication-driven cognitive engagement, exploration-oriented regulation and high-level socio-emotional expression, with the medium programming task score. Based on these findings, pedagogical and analytical implications were proposed to promote CSRACP in higher education.

The experience of loading Web Apps has gained a lot of attention from both academic researchers and industrial practitioners. State-of-the-art efforts mainly focus on reducing the metric of Page Loading Time (PLT), making the resources of Web Apps retrieved from the network and evaluated on the browser as soon as possible. However, our user study on comparing the experience of loading 1064 popular Web Apps shows that shorter PLT does not imply better experience in about 25% comparison pairs. Then, for the Web Apps in these pairs, we use seven visual metrics to measure the loading experience and find that visual factors dominate the experience. To this end, we summarize four optimising principles from the perspective of visual factors to improve the experience of loading Web Apps. These principles are easy to be adopted by Web App developers and can complement the efforts of reducing PLT.

On-device learning is a promising technique for emerging privacy-preserving machine learning paradigms. However, through quantitative experiments, we find that commodity mobile devices cannot well support state-of-the-art DNN training with a large enough batch size, due to the limited local memory capacity. To fill the gap, we propose Melon, a memory-friendly on-device learning framework that enables the training tasks with large batch size beyond the physical memory capacity. Melon judiciously retrofits existing memory saving techniques to fit into resource-constrained mobile devices, i.e., recomputation and micro-batch. Melon further incorporates novel techniques to deal with the high memory fragmentation and memory adaptation. We implement and evaluate Melon with various typical DNN models on commodity mobile devices. The results show that Melon can achieve up to 4.33× larger batch size under the same memory budget. Given the same batch size, Melon achieves 1.89× on average (up to 4.01×) higher training throughput, and saves up to 49.43% energy compared to competitive alternatives. Furthermore, Melon reduces 78.59% computation on average in terms of memory budget adaptation.

The rise of metaverse has driven the burst of 3D Web apps to provide immersive experience across heterogeneous devices. However, loading 3D models in Web apps is usually slow especially on mobile devices with limited computation capability and dynamic network condition, harming the experience of 3D mobile Web apps. In this paper, we propose an approach called 3Dispatcher to reducing the 3D model loading time by adaptively compressing the 3D models. Preliminary results show that 3Dispatcher can speed up the 3D model load by 15%-41% in different scenarios.