High-quality assessment at scale requires questions that are conceptually rich, appropriately difficult, and cognitively engaging. We present a learning analytics pipeline that transforms archival assessments and learner performance data for question generation. Past assessment questions from an undergraduate engineering physics course are multi-labelled with topics from a predefined syllabus, and a primary large language model (LLM) is selected via inter-rater reliability against expert judgments to ensure label quality. Using these labels and student performance data, we construct an undirected weighted graph where nodes represent topics and edges capture both co-occurrence and difficulty, with student performance used as a proxy. Exploring this graph reveals difficulty-calibrated topic sets that can be retrieved around target topics. We then evaluate LLM-generated questions for difficulty and alignment with the Structure of Observed Learning Outcomes taxonomy when provided with the requested topics. Empirically, we found that GPT-5 Thinking shows higher agreement with expert labels. We further found that the generated questions, grounded by the topic-performance graph, were deemed to be more difficult and of higher-order intent under blinded expert review. The contribution is a scalable creation of higher-order assessment items with appropriate difficulty that are imperative for generative personalised learning and assessment systems.

Learning at scale often requires domain-specific automation such as assessment and feedback. An organization locked in to a general learning platform without these specialist automations limits its pedagogical offering. An ecosystem of interoperable, platform-agnostic microservices for domain-specific automation would solve this problem. To develop an effective ecosystem, a standard interface (API) for education microservices is required.

Predicting examination grades is essential to facilitate early interventions and to enhance student retention rates in an academic institution. We propose a predictive model based solely on historical academic performance made available before the beginning of each semester. The proposed model employs singular value decomposition to distill the underlying student-course graph structure, resulting in a student representation vector that holistically captures a student’s academic ability across courses in relation to their cohort. This representation vector is then fused with the student’s historical academic records for grade prediction. Data for training the proposed prediction model was sourced from approximately five thousand Electrical and Electronic Engineering students across seventeen core courses, including Circuit Analysis and Analog Electronics taught in the sophomore year. Students identified as at risk of failing a course at the beginning of each semester may be offered targeted (academic) support such as peer tutoring programs.

The need to cultivate creativity in engineering education calls for opportunities for students to exercise freedom in proposing and pursuing projects aligned with their interests. This paper presents insights into an undergraduate makerspace-based course in terms of factors affecting students’ creativity-related self-efficacy. We conducted a survey on students who come from an engineering and science background to gain their opinions about the impact of this course on enhancing their creativity. To establish if there is a significant difference in the students’ creativity, we performed the non-parametric Wilcoxon signed-rank test comparing the first and second survey, with results showing that there is a statistically significant increase in students’ creativity-related self-efficacy. There was a general increase for all the items, especially students’ perceptions toward the relevance of the course, the conduciveness of the learning environment, opportunities to make and learn from mistakes, and the resourcefulness of their team. Results obtained via quantitative statistical analysis was backed up by qualitative analysis that employed text mining techniques such as automatic key phrase extraction and sentiment analysis on the open-ended responses and the reason(s) for the Likert-scale answer choice. In addition, we used the Spearman’s rho to report correlations between Likert-scale items and determine the variables that are significantly and positively correlated with the creativity-related self-efficacy construct. A multivariate regression model was then constructed to observe the extent to which each highly correlated variable impacts creativity-related self-efficacy; of which, a sense of relevance appears to have the largest effect. Through gaining insights into the factors that may impact students’ creativity-related self-efficacy, this study contributed to a deeper understanding on how this important attribute could be developed through a makerspace-based university course.

This Research Full Paper presents findings from a quasi-experimental study on the effectiveness of collaborative learning in improving theoretical understanding within a large electrical engineering laboratory course. While there are several studies that report positive student responses and performance gains with collaborative learning in higher education STEM laboratories, as far as we know few studies include control groups and prepost-test to ascertain the impact of such approaches on students’ perception and performance gains.142 students were surveyed on their perceived theoretical understanding under three experimental settings: an experimental group involving 90 students in a magnetic fields laboratory with collaborative learning (E), a control group involving the same students in an electric fields laboratory (C1) and a second control group involving a different group of 52 students who went through the same laboratory as E without collaborative learning (C2). The study found that students’ perceived theoretical understanding in E improved significantly before and after collaborative learning (d=1.05). Significant difference in perceived theoretical understanding between E and C1 was found but this result needs to be interpreted with caution as students perceived C1 to be significantly more difficult. Nevertheless, students in C1 also feedback that collaborative learning would have aided their theoretical understanding. Furthermore, students’ perceived theoretical understanding in E was significantly higher than those in C2 when laboratory sequence was taken into account (d=0.394), suggesting that collaborative learning was helpful for theoretical development. The scores achieved by a subset of these students in the E (n=69) and C2 (n=20) setting were also compared. A non-significant increase in performance were found with E, suggesting that short-term performance gains might be limited.The qualitative feedback from students suggest that knowing the responses of their peer enabled them to understand the concepts from many perspectives and identify their mistakes. Furthermore, peer discussion provided affirmative feedback to some students which helped them gain confidence in their understanding. Hence, it is prudent for the study to conclude that including collaborative learning in engineering laboratory has provided students with a sense of much needed support with no negative consequence on theoretical knowledge gain.

This Innovative Practice Full Paper presents experiences in designing a student-led virtual freshmen orientation program that uses a Minecraft environment. We describe the planning process, roles of the organizing committee members, and how the game was constructed for participants to learn and interact with one another. The student organizers not only created a virtual environment that scales the college map where more than a hundred freshmen (participants) could have an immersive experience of the campus, but also ensured the branding and marketing, logistics, and safety/well-being aspects of the event. In this paper, we present students’ experience of this program from both the designers’ as well as the participants’ perspectives. We conducted surveys with the organizing committee members and interviewed the participants to gain insights on their perception of this event. Our analysis showed that student organizers had the autonomy to brainstorm, suggest creative ideas, develop novel games, and procure materials. They also felt that they developed authentic programming and leadership skills. On the other hand, participants felt engaged as the event was well-organized, had clear delivery of information, introduced them to new technology, made them more familiar with the campus, provided a conducive environment to hone their soft skills such as communication and teamwork even before they officially enrolled as undergraduate students in an engineering program, and helped them establish social networks to support them throughout their undergraduate education journey.

The paper proposes a fuzzy image labeling method that assigns multiple semantic labels together with confidence measures to each region in an image. The confidence measures are derived from the distance of the region to hyperplanes constructed by support vector machines. Test results show that this method yields higher classification accuracy and retrieval precision than crisp labeling methods that are based on crisp classification.

Histogram-based dissimilarity measures are extensively used for content-based image retrieval. In an earlier paper, we proposed an efficient weighted correlation dissimilarity measure for adaptive-binning color histograms. Compared to existing fixed-binning histograms and dissimilarity measures, adaptive histograms together with weighted correlation produce the best overall performance in terms of high accuracy, small number of bins, no empty bin, and efficient computation for image classification and retrieval. This paper follows up on the study of adaptive histograms by applying them to texture classification, retrieval, and clustering. Adaptive histograms are generated from the amplitude of the discrete Fourier transform of images. Extensive comparisons with well-known texture features and dissimilarity measures show that, again, adaptive histograms and weighted correlation produce good overall performance.