Supporting Group Coursework Assessment in Large Computing Classes through an Open-Source Web Application

Group assessment in computing courses is widely recognised as problematic, particularly in large cohorts. Persistent issues include uneven workload distribution, poor communication, and limited progress monitoring. Existing practices for group formation, peer review and meeting tracking are manual and disjointed. We explain the importance of addressing these issues (section 2), and related work on group allocation and peer reviews (section 3). Our contribution builds on this prior research by adapting such ideas to a large institutional context, embedding them into curricula, and providing an open-source tool designed for practical deployment, presented in section 4. We also describe its context of application and a provide preliminary evaluation (section 5). We then offer some conclusions and future work (section 6).

Soft interpersonal skills such as teamwork and communication are key employability factors [15] that may be seen as more important than the technical skills acquired at university [28]. Evidence suggests both are equally influential in determining salaries [3].

Many approaches exist for teaching soft skills in higher education, though group-based projects are common [9]. These can take many forms, ranging from informal seminar groups to year-long projects contributing a significant proportion of marks. Implementations across universities, subjects and individual modules vary wildly, and despite recommendations of best practice, there is little standardisation in how such activities are organised [11].

Beyond simply allowing students to practice interpersonal skills, group work offers other benefits, such as helping universities to manage rapidly growing cohorts [14, 21].

Despite its clear advantages, group work in higher education has many challenges. Most notable is social loafing, the reduced effort someone exhibits when working in a group versus working alone [18]. Social loafing continues to be observed in group-based assignments and can lead to conflict within the group [23]. The effect is known as free-riding, though some argue these are distinct, with free-riding being a consequence of social loafing and involving an active desire to gain credit without working [14].

A related phenomenon is the sucker effect: a form of social loafing that occurs when a group member believes other members lack effort. That member then reduces their own contribution to match what they perceive the effort level of the group to be [27]. Empirical evidence has shown it at work even in projects designed to involve the participants personally [26]. Together, these behaviors lead to dissatisfaction with performance within groups, which, in turn, causes frustration around the final grades [1]. Still, such experiences provide valuable learning experiences for students [6].

Recent work [2] highlights current challenges around fairness, equality, and student attitudes toward teamwork in computing education, reinforcing the need for supportive tools that encourage positive collaboration while addressing perceived inequities. This adds to a body of work on collaborative learning and the use of systems for peer assessment, team allocation, and workload tracking. Existing studies demonstrate the potential of algorithmic approaches to support group formation and peer evaluations.

Our contribution builds on this prior work by unifying these tools into a single system, facilitating a fairer assessment, better oversight, and earlier intervention when teams struggle. We developed an open-source web application (github.com/samuelkitson/group-coursework) to support group coursework management in undergraduate computing. Key features include:

• computerised team allocation using genetic algorithms;
  
• meeting tracking for accountability;
  
• structured peer reviews and skills self-assessments;
  
• automated insights about struggling students/teams;
  
• supervisor dashboards and tools.
  

The system replaces informal, ad-hoc processes with structured, data-driven support for students, staff and team supervisors at each stage of the coursework.

4.3 Peer reviews

Drawing upon the work described previously, we recognised the importance of regular, low-stakes peer reviews in providing evidence for persistent free-riding. Our approach, called check-ins, is to ask students to rate the relative workload in their team every week during the project, considering only their experiences in the preceding week. This aims to limit the effect of short-term problems (such as illness) on a student's overall scores; as data is gathered weekly, only their scores for that week will be affected and this will be easy for staff to identify.

Care must be taken when designing the user interface for such a tool; some student interviewees admitted that when using existing points-based systems (e.g. [8]), they inflate their own ratings without fully considering the contributions of others. To combat this, the check-in asks students to distribute effort points among the team (between one and seven each) to reflect perceived contributions. Each member starts with four points and a zero-sum constraint prevents points being added without being taken from somebody else (Figure 1). This encourages consideration about relative contributions and whether the student had to complete another's work. A similar approach has previously been effective [19].

Staff can also configure full check-ins for specific weeks, which ask students to rate each other in specific skills and provide a short review comment. These could inform marking or be released to students as appropriate for the assignment.

4.4 Insights and reports

The key benefit of combining the above functionality into a unified system is that the data is held centrally and can be used to generate detailed insights about student progress. This allows struggling students or teams to be flagged to staff, who can intervene and provide support before serious problems develop.

Various data points (such as meeting attendance and check-in scores) are analysed to generated insights about each team and made visible to staff and supervisors. For convenience, these are categorised by severity to allow teams with the most serious problems to be easily identified.

Detailed progress reports can be generated about teams to inform marking. Among other statistics, they compare students’ check-in scores for themselves compared to those received from others (Figure 2). This makes it possible for staff to deduct marks or investigate further when a student's self-reported effort differs significantly from the team's perception; an indicator of free-riding.

The tool has been embedded into COMP2300, Software Design and Development Project, a year-long second-year undergraduate module, with 300 students in 2025/26. The module is project-based, requiring sustained collaboration across multiple deliverables. The application is integrated into students’ project workflow, and used by staff to monitor progress and provide targeted support. Initial trials within the department indicate promise. Students are optimistic that the tool could lead to a less uneven workload and a more enjoyable experience. Staff report that it helps identify struggling teams earlier than before and provides evidence for fairer marking. System data (i.e., longitudinal peer reviews, supervisor observations and meeting logs) offer richer insight into team dynamics.

A short initial study was conducted with 55 student participants, who completed an online questionnaire to understand their perceptions of group coursework. Three months later, and after a brief introduction to the app, 22 participants returned to access a personalised account with two example assignments. To mimic everyday interactions with the app, they were asked to complete a skills assessment, record minutes for a simulated meeting, answer a check-in and lodge a meeting dispute.

Nine Likert questions from the initial study were then repeated to identify changes, with the statements adjusted to ask how participants would feel if using the app for a real coursework. The Wilcoxon signed-rank test was applied to the results and the significance value of 0.05 reduced to 0.00556 using the Bonferroni correction. Three statements saw statistically significant changes, each becoming more positive after exposure to the app (Figure 3).

Free text responses from the questionnaire and comments from a student focus group provide insight into these improvements. A recurring theme was that students felt reassured that the app gathers clear evidence in the event of free-riding, leading to increased accountability. One participant said it was:

“An exceptionally useful tool for holding people accountable for the work they promised to do.”

The meeting tracker was highlighted as effective for generating this evidence and simplifying the process of recording minutes. One student raised concerns about the integrity of meeting minutes if they can be edited following a dispute; in response, we implemented an audit log to track changes.

The check-in feature was well-received. Confidence in the ability of regular peer reviews to encourage equal effort increased after students completed the simulated check-in. Many participants commented positively on the zero-sum constraint, for example:

“Balancing points makes you think about the distribution of the workload more deeply since the setup stops you deducting points without reallocating them.”

Some participants were concerned that the simple numeric check-ins would be ineffective when no other team members had contributed, though this would be mitigated by staff enabling full check-ins and requiring students to provide a review comment.

Formal evaluation with the first full cohort is ongoing. Planned analysis includes comparisons between self- and peer-reviewed scores, and correlations between generated insights (e.g. attendance, review comment quality, etc) and project performance.

This work demonstrates how structured, data-driven support for group work can be scaled to large undergraduate cohorts. It offers practical lessons for educators: which features are most useful, what data can realistically be collected, and how to embed such tools into existing assessment structures. As the system is open source, it has potential to be adopted and adapted by others in the computing education community facing similar challenges.

Our future work includes completing evaluation of the current cohort, disseminating results, and refining the tool. Planned enhancements include considering peer review data from previous semesters during allocation, additional feedback mechanisms, and integration with institutional learning management systems. We also aim to trial the system in other modules and disciplines, exploring scalability and transferability beyond computing.