The rapid adoption of LLMs in higher education (HE) has raised critical questions about their effectiveness in feedback provision. This systematic scoping review explores key trends, research gaps, and the pedagogical alignment of LLM-generated feedback. The findings highlight the growing integration of LLMs in assessment practices, their potential to improve feedback quality, and the need for pedagogically structured implementation. This review offers valuable insights for researchers, educators, and institutions, supporting the strategic adoption of generative AI to optimize feedback systems and improve student learning outcomes.

Allocating students to projects is a commonplace task in computing education. These decisions underpin student-supervisor allocation, the formation of tutee and capstone groups, and pair programming. These allocations play a critical role for individual learner outcomes and the success of collaborative interventions. For example, imbalance in either gender, ethnicity, or nationality can negatively impact learner outcomes. Despite the critical importance of these allocation choices, we see little consensus on how these are implemented. The allocation task can be challenging and time-consuming for instructors of even moderately-sized classes, and the fairness implications can be difficult to assess. Inadvertently, an instructor may allocate in a way that amplifies existing biases or disproportionately harms those from disadvantaged or protected groups. From students' perspectives, a lack of transparency on the allocation process may also lead to issues of trust. The Working Group will undertake a study of allocation practices by bringing together educational and ML literature to develop and evaluate the fairness of allocation methods, and develop educator guidelines to promote pedagogically grounded allocation practices.

This paper presents a new category that has been added to the classification of Kim and Ko (2017) for programming learning systems, namely the Online Coding Tutorial System (OCTS) category. In this current study, firstly, seven popular online coding tutorial systems have been selected to investigate how these systems taught learners and what their characteristics and features were. Secondly, from Kim and Ko's classification, one system has been selected from each category and analyzed across the identified characteristics of online coding tutorial systems to investigate whether any existing category in Kim and Ko's classification shares the same characteristics. As a result, it was found that online coding tutorial systems have adopted many of the features that have been identified in Kim and Ko's first category of interactive platforms, along with some aspects of their creative platforms and MOOCs. Therefore, online coding tutorial systems have been considered a new category of programming learning systems that includes several characteristics from other existing categories.

Enhancing student engagement in computer science (CS) courses is crucial to improving students' achievement and fostering active participation in computer science education (CSE). Previous studies have highlighted different factors that shape student engagement, including behavioural, cognitive, emotional, and social engagement. Additionally, other factors influence student engagement, such as students' beliefs in the usefulness of learning computer science and their confidence in taking CS classes. Despite existing studies investigating student engagement in CSE, limited studies have explored factors that influence novice student engagement in CS courses. Further, no study has applied machine learning (ML) techniques on the combined self-reported engagement data and student log data to predict CS learning performance. Therefore, this study used ML techniques to explore and identify student engagement factors that affect and predict CS learning outcomes. To achieve this, data was collected using three different sources: self-reported data, system log data, and CS learning performance data. Log data from student behaviour on the system included monthly logs of student access to the CS course page on the LMS during the semester, the total hits of student interactions with the course content throughout the semester, and the total number of task submissions. Our analysis involves 77 novice students who consented and completed a multidimensional self-reported questionnaire during the second semester of 2022 - 2023 at a university in Saudi Arabia. The K-means clustering algorithm was used to understand engagement patterns by classifying students into groups based on their levels of self-reported engagement, log data, and academic performance. Classification algorithms using Random Forest (RF), Decision Tree (DT), and LightGBM (LGBM) were used to predict CS learning performance from student engagement data (self-reported and logs). We evaluated the performance of ML algorithms using metrics including accuracy, precision, recall, and Fl-score. The clustering results showed that students who actively engage with the course content (log data) tend to achieve higher grades, especially those with higher total hits of student interactions with the course content throughout the semester. The classification results showed that the RF model outperforms DT and LGBM, highlighting the significance of student interactions with the course in the first month of the semester as the key indicator influencing CS learning performance. Our findings contribute to a deeper understanding of student engagement in CS education and highlight various sources and factors used to measure and influence student engagement. The study has implications for educators, researchers, and stakeholders who may design effective interventions that would increase engagement to improve student learning outcomes in CS education. Future work will use a larger sample of participants from various educational levels in different countries.

This paper explores the potential of ePortfolios in supporting students’ transition through their first semester in the School of Computing Science at a Scottish university. It examines the relationship between ePortfolios use, mentoring, reflection on writing and recording, communication skills, and critical reflection among international students and those joining university straight from colleges (further education) or high schools (n = 75). Results suggest that effective mentoring, engagement in reflective practices, and development of communication skills can facilitate a smoother transition through higher education and enhance students9perceived benefits of using ePortfolios.

We present the first prototype of Curio, an affordable robot designed for computer science education (CSE). Curio’s goal is to enhance students’ engagement and facilitate CSE. This is achieved by placing students’ smartphones at the center of their experience and letting them designing standalone activities that run on web pages. Such web applications can access all smartphone sensors, including the front and back cameras, and control the robot directly via Web-Bluetooth. This paper provides an overview of Curio’s design philosophy and the technological integration of our first prototype. Videos of Curio in action are available at trycurio.com.

Females remain underrepresented in computer science (CS), despite numerous studies investigating the causes using different data types and analysis techniques. In recent years, machine learning (ML) has been increasingly used in education, particularly for analysing student engagement using some engagement dimensions. However, no study has yet used ML algorithms to analyse student behavioural, cognitive, emotional, and social engagement survey data. In this paper, we present a study investigating whether these four dimensions of engagement are related to female high school students' beliefs in the usefulness of learning computer science and their confidence in taking CS classes in Saudi Arabia. We also employ ML techniques to identify important indicators that can predict students' confidence and beliefs in the usefulness of learning CS. Additionally, we compare three supervised ML techniques, Random Forest (RF), Decision Tree (DT), and LightGBM (LGBM), to determine which algorithms better predict confidence and usefulness in learning CS. Our sample consisted of 284 participants from four schools who completed the multidimensional survey, and we evaluated the ML algorithms using Mean squared error (MSE), Mean absolute error (MAE), and Determination coefficient (R2). Our findings show that each dimension of student engagement positively correlates with the confidence and usefulness of learning computer science, and it is possible to predict them from student engagement indicators. The RF model outperformed DT and LGBM, identifying 'enjoyment in learning new things' and 'interest in topics in CS class' out of 28 indicators as the most important features to predict usefulness. Additionally, from 28 features, 'looking forward to CS class' and 'enjoyment in learning new things' are the most important indicators influencing confidence. Our findings contribute to a broader understanding of student engagement in CS education and highlight various indicators used to measure student engagement. These findings shed light on factors that may motivate and interest female students toward CS learning. Future work will use a larger sample of participants from schools and higher education in different countries.

This paper presents a dual dashboard early warning system which uses students' affective state as a measure of risk. Affective state has been shown to influence CS1 performance, and specific states such as frustration have been linked to attrition. The software administers affective surveys to students using a series of 2-dimensional grids. Students then complete a qualitative journal entry. Risk weights are assigned to students based on the journal response's sentiment analysis and whether student's 2-dimensional grid responses fall within configurable 'danger zone' bounds. The early warning system automatically flags students as needing support if the responses' combined risk weights exceed configurable thresholds. Additionally, flags can be assigned manually, either by instructors or by students themselves.

Full Paper. Research. We discuss possible ways to direct students to right level of introductory programming. While many schools offer college preparatory or advanced placement courses in computing, there is still, unfortunately, a large part of the "college-ready" population that has no opportunity to learn computing at all before they arrive. Regulation of CS education at the state/province or national level is still rare (but growing). Thus incoming students possess a wide range of skills and knowledge. When coupled with increasing enrollments, this diversity of experience can result in courses having large numbers of both absolute beginners and seasoned coders. Such courses are difficult to teach, intimidate novice students, and bore those with more experience. This can result in low engagement and retention.Unlike mathematics and language arts, introductory courses in CS vary widely from one institution to another in both conceptual material and programming language used. A standard point of entry to college mathematics is a calculus course, with some students instead starting earlier with pre-calculus or an algebra refresher, and others starting out in the second-term calculus course. There is rarely a concern about student skill being hidden by notational or other language differences, because the language of mathematics is close to universal. Similarly, freshman language arts courses in reading and/or writing assume a certain level of skill and maturity of comprehension and expressiveness in the target language; otherwise remedial courses are provided.We investigate placement of incoming first year students into appropriate introductory computer science courses at higher education institutions where there is more than one choice of first course. The goal is to determine the best way to decide which first course would be the most helpful for each student.

This paper describes the development of `StudentsAtRisk', a prototype early detection system. It is based on engagement with course materials and can be used to identify students who are falling behind by automatically flagging them. The system also allows instructors to flag these students manually. On their dashboard, students can flag themselves, as engagement with the material might not reveal all those who struggle.

Goal-directed problem-solving labs can lead a student to believe that the most important achievement in a first programming course is to get programs working. This is counter to research indicating that code comprehension is an important developmental step for novice programmers. We observed this in our own CS-0 introductory programming course, and furthermore, that students weren't making the connection between code comprehension in labs and a final examination that required solutions to pencil-and-paper comprehension and writing exercises, where sound understanding of programming concepts is essential. Realising these deficiencies late in our course, we put on three 3-hour optional revision evenings just days before the exam. Based on a mastery learning philosophy, students were expected to work through a bank of around 200 pencil-and-paper exercises. By comparison with a machine-based hackathon, we called this a Thinkathon. Students completed a pre and post questionnaire about their experience of the Thinkathon. While we find that Thinkathon attendance positively influences final grades, we believe our reflection on the overall experience is of greater value. We report that: respected methods for developing code comprehension may not be enough on their own; novices must exercise their developing skills away from machines; and there are social learning outcomes in programming courses, currently implicit, that we should make explicit.

We propose a working group to investigate methods of proper placement of university entrance-level students into introductory computer science courses. The main issues are the following. The ability to predict skill in the absence of prior experience The value of programming language neutrality in an assessment instrument Stigma and other perception issues associated with students' performance, especially among groups underrepresented in computer science The impact or potential impact on underrepresented populations (minorities, those with lower socioeconomic status) The outcomes/satisfaction/retention metrics in the major of the paced/tracked students compared to those in one-size-fits-all introductory classes

Assessment feedback, which is an important factor in student learning development, has been a source of dissatisfaction for many years. The technology used in universities such as Gradebook does not allow the tracking of student engagement with their feedback. In general, engagement is mainly reported via observation or students' self-report. This paper presents results from using a prototype application to track students' engagement with their feedback in three studies (n=218, n=70 and n=148) involving summative and formative assessment feedback. Using digital footprinting data, it was possible to gather detailed information on students' access and engagement with their feedback.