Marking group work fairly

Group assignments produce the most heated complaints in many units, and the complaint is nearly always the same: "I did most of the work and got the same mark as someone who did nothing." Giving everyone the group mark is simple but often unfair, while trying to assess every individual contribution can become unmanageable. This guide shows how to design the marking so that the shared product and individual contributions are both assessed, with a worked example of how the marks combine.

Decide what the group work is for

Before choosing a marking scheme, be clear about why the task is a group task. The answer shapes everything else.

• Collaboration is a learning outcome. Teamwork, negotiation and project management are being assessed. You need some evidence of how students worked together, not just what they produced.
• The task is too large for one person. A full software system, a field study or a major design project. Here the product matters most, and the question is whether each student contributed fairly.
• It saves marking time. Honest, and common. But if the task is a group task only for logistics, keep the group component modest and assess the learning outcomes individually wherever you can.

Separate group criteria from individual criteria

The most reliable way to reduce unfairness is to assess some things at group level and some individually, rather than trying to adjust a single group mark afterwards.

A common split is 50 to 70 percent group, 30 to 50 percent individual. The individual component should test something that genuinely needed the student to engage with the project. A 300 word reflection written the night before tests little; a five minute individual viva where each student explains a design decision tests a great deal.

Peer assessment

Peer assessment asks group members to rate each other's contribution. Used carefully, it gives you information you cannot get any other way. Used carelessly, it becomes a popularity contest or a tool for settling scores.

Designing the peer rating

• Rate specific behaviours, not overall worth. For example: attended and prepared for meetings; completed agreed tasks on time; contributed ideas; helped integrate others' work; communicated when problems arose. Five items on a 1 to 5 scale is plenty.
• Include self assessment. Large gaps between self and peer ratings are useful to look at, in either direction.
• Ask for a short justification for any rating of 1 or 2. This discourages careless low ratings and gives you something to investigate.
• Keep ratings confidential from other group members, and tell students that you will look at the evidence behind unusual ratings rather than applying them automatically.

Turning ratings into marks

A widely used approach is a contribution factor. Each student's average peer rating is divided by the group's average, giving a factor around 1. The factor then scales the group mark.

Set limits on the factor so that it cannot swing marks wildly. For example, a floor of 0.5 and a ceiling of 1.1 means a student rated as contributing little can lose up to half the group mark, while even the strongest contributor gains no more than 10 percent. Choose the limits before the task starts and publish them.

Some coordinators only apply the factor when it falls outside a band, such as 0.9 to 1.1, and treat anything inside it as normal variation. This avoids small, noisy adjustments that students find hard to accept.

A worked split of marks

A third year engineering design project, four students, total 100 marks:

• Group component (60 percent): design report and prototype, marked with a rubric. The group earns 70 out of 100.
• Individual component (40 percent): a ten minute individual viva on the design, marked out of 100.
• Peer contribution factor: applied to the group component only, limited to between 0.5 and 1.1.

Peer ratings (total of five items, each rated 1 to 5 by the other three members, then averaged):

Final marks:

Before applying Dylan's factor, the coordinator checks the evidence. The shared repository shows no commits from Dylan after week 6, meeting notes record three missed meetings, and Dylan's own self assessment acknowledges falling behind. The viva shows a reasonable understanding of the early design but not the later work. The adjustment is supported, and the coordinator records the evidence in case of appeal.

Had the evidence been thin, for example if Dylan's low rating came from one team member and the commit history showed steady work, the coordinator would investigate further or not apply the factor.

Note that without the individual component, Aisha and Chen would have received the same mark despite a clear difference in how well they could explain the design. And without the peer factor, Dylan would have passed comfortably on the strength of the others' work.

Collecting contribution evidence

Peer ratings are more trustworthy when they can be checked against something. Useful evidence, depending on the discipline:

• Version history. Shared documents and code repositories record who changed what and when. Treat them as a signal, not a measure: one person may commit everyone's work, and a large commit may be formatting.
• Meeting records. A brief template (date, attendees, decisions, actions and who owns them) completed each week.
• Task allocation. A simple table at the start of the project showing who is responsible for what, updated if it changes.
• Checkpoints. A short progress check with the tutor at one or two points during the project. Problems raised in week 5 are easier to fix than those raised after submission.

Tell students at the start what evidence you may look at, and why. Knowing that contribution will be visible changes behaviour more than any penalty.

Handling group breakdowns

Some groups will fail to work together however well you design the task. Plan for it:

• Give students a way to raise problems early, such as a mid-project check-in where each student privately reports on how the group is going.
• Set out what happens if a member stops contributing. Can the group reassign their part? Will the non-contributor be moved or given an individual task?
• Do not wait for the final peer assessment to act on a problem you already know about.

Communicating the scheme

Publish the weighting, the peer assessment items, the factor limits and a worked example in the assessment brief. Students accept adjustments they understood in advance far more readily than ones they discover when marks are released. When you release marks, show each student their group mark, their factor and their individual mark separately, so the final number is not a mystery.