TL;DR: A group project can provide evidence about a shared product, the process of collaboration, and each student's understanding. Those are different claims and should have explicit criteria and evidence. Announce the grading arrangement before work begins, create opportunities for individual demonstration, and treat peer reports as information requiring interpretation. A strong team output should not automatically certify every member's learning, while an uneven contribution should not be inferred from visibility or popularity alone.
The contract before the project
A fictional urban planning course asks teams of four to propose an improvement to a neighborhood transport interchange. Each team submits a recommendation, a supporting analysis, and a presentation. The instructor wants students to learn both policy reasoning and collaboration. In the previous version of the assignment, every team member received the same grade for the final product.
That arrangement made the grading simple but left an evidentiary gap. A polished proposal could be produced through unequal understanding, effective specialization, extensive outside help, or genuine shared learning. The final artifact alone did not distinguish those possibilities. The instructor therefore redesigns the assessment before the next project begins.
The cases in this article are original fictional examples. The project, students, marks, and events are invented to illustrate assessment decisions. The proposed grading contract is not a universally validated formula. It shows how an instructor can make the relationship between evidence and grade explicit, then examine difficult cases without inventing new rules after the work is complete.
Carnegie Mellon's guidance on assessing group work distinguishes group products and processes from individual contributions and understanding. It also cautions that peer reports are not foolproof. That distinction is the foundation for the contract: the course should state which claim each assessment component supports.
Three claims, three evidence routes
The first claim concerns the shared product. Does the proposal define a relevant problem, compare plausible options, use evidence responsibly, and justify its recommendation? The team artifact can directly support this judgment. The criteria apply to the work submitted, regardless of which member wrote a particular paragraph.
The second claim concerns individual understanding. Can each student explain a central decision, interpret the evidence, and respond to a changed condition? This requires an individual demonstration. It might be a short written analysis, a structured conversation, or another appropriate task aligned with the learning goals and approved access arrangements.
The third claim concerns collaboration. Did the student fulfill agreed responsibilities, communicate dependencies, contribute to decisions, and respond constructively when the work changed? This requires evidence of the process, not merely the final result. Peer reports may contribute, but they should be connected to observed behavior and considered alongside other information.
The instructor assigns weights of fifty percent to the shared product, thirty percent to individual understanding, and twenty percent to collaboration. These are fictional design choices for this course. Another course may reasonably use different weights or avoid combining the components into one score. The essential requirement is to explain the intended meaning of the weights before students organize their work.
The calculation should reveal its assumptions
Suppose a student receives eighty out of one hundred for the team product, seventy for individual understanding, and ninety for collaboration. The weighted result is forty plus twenty-one plus eighteen, giving seventy-nine out of one hundred. The arithmetic is transparent, but the quality of the grade still depends on the evidence behind each component.
A weighting formula cannot rescue a poorly defined collaboration score. If that score reflects popularity or how often someone spoke, the total will combine a defensible product judgment with an indefensible process judgment. The instructor needs observable criteria and an opportunity for students to understand and respond to relevant concerns.
The contract also explains whether any minimum individual-learning requirement applies. In this fictional version, the course reports all three components and uses the published weighted total; it does not introduce an unannounced threshold after marking. If a course requires a minimum demonstration of a particular capability, that condition should be explicit and consistent with institutional rules from the beginning.
Carnegie Mellon's comparison of group grading methods illustrates that different methods create different tradeoffs. The formula here is an original example. It should be judged by what the course wants the grade to mean, rather than adopted because a set of percentages appears balanced.
Students need a way to distribute work without distributing learning away
The team must divide tasks. Requiring every student to perform every operation would make collaboration inefficient and may defeat the purpose of the project. Yet specialization can become a learning problem if one student alone understands the analysis while another merely formats slides.
The assignment therefore identifies a common core that every student must understand and areas where specialization is acceptable. Everyone must explain the main recommendation, the central evidence, and one material limitation. Students may specialize in data preparation, stakeholder analysis, visual communication, or another defined contribution. The individual demonstration focuses on the common core and the student's own area of responsibility.
A brief midpoint exchange helps students teach one another. Each member explains their component to the group and identifies a decision that depends on someone else's work. The aim is not to require a ceremonial presentation. It is to expose dependencies early enough that the project can change and students can learn from one another.
The instructor provides time and expectations for that exchange. If shared understanding is part of the grade but the schedule rewards only rapid assembly of separate pieces, the design creates conflicting incentives. Students should not be penalized for failing to infer a collaboration method that the assignment neither taught nor made feasible.
Case one: the invisible contribution
The first team submits a strong proposal. Three members say that Hana “did not contribute much” because she spoke less during meetings and did not present. Hana's work record shows that she cleaned the dataset, identified inconsistent definitions, and corrected a calculation that changed the recommendation.
The instructor should not treat either the peer majority or Hana's self-report as an automatic verdict. The relevant questions are what responsibilities were agreed, what work occurred, how it affected the project, and whether the collaboration criteria were met. Speaking frequency is not the same as contribution, although communication may still matter when others need to understand the work.
Hana provides the relevant version history and explains the correction. The other students acknowledge that the analysis was useful but say they did not understand its status until late. The evidence now supports a more precise judgment: substantive analytical contribution alongside a communication problem. Reducing the entire case to either “free rider” or “unrecognized hero” would lose that distinction.
The instructor evaluates the two aspects under the published criteria. Hana should receive credit for the work she performed. If she failed to communicate a critical dependency in a timely way, that can also be addressed when the assignment established such an expectation and she had a fair opportunity to meet it. The purpose is a proportionate assessment of behavior, not a character label.
The case also reveals a design issue. The team had no routine way to make behind-the-scenes work visible. A short shared task record or midpoint demonstration could have prevented some misunderstanding. The instructor can improve that process without requiring students to produce an exhaustive surveillance log of every minute they worked.
Case two: the polished product and the weak individual explanation
A second team's proposal receives eighty-eight for the product. During an individual task, one member cannot explain why the chosen option was preferred or what evidence would weaken the recommendation. The student can describe the slide layout but not the central argument.
The instructor should first check whether the individual task assessed the announced common core under appropriate conditions. Was the question clear? Did the student have the agreed access support? Was the expected level of detail consistent with the teaching? A poor response is evidence to interpret, not permission to assume laziness or dishonesty.
If the task was appropriate and the gap remains, the individual-understanding component should reflect it. The strong group product does not establish that this student mastered the reasoning. At the same time, the instructor should not automatically remove all credit for the shared product if the grading contract says that component is shared.
Suppose the student receives forty-five for individual understanding and seventy for collaboration. Using the announced weights, the total is forty-four plus thirteen and a half plus fourteen, or seventy-one and a half. The result may prompt a debate about whether the weights express the course's priorities. That debate belongs in course design and future revision, not in an improvised penalty for this student.
Where institutional policy permits reassessment or additional learning opportunities, the instructor can use the observed gap to guide them. The opportunity should follow a clear rule rather than depend on how persuasive the student is in requesting an exception. Assessment can reveal a learning need without pretending that the final product already satisfied it.
Case three: the member who did everything
A third team produces a technically excellent proposal, but one student, Leo, rewrote most of the work and excluded others from important decisions. Leo argues that the grade should reflect his extraordinary effort. The other students say they were unable to contribute meaningfully after their drafts were repeatedly replaced.
The instructor needs to distinguish high-quality production from collaboration. If the course explicitly assesses working with others, taking over may undermine a learning objective even when it improves the artifact. That does not mean every act of extensive revision is wrong. The relevant evidence concerns agreements, opportunities, communication, and the reasons changes were made.
Perhaps the team missed deadlines and Leo responded to a genuine risk. Perhaps he rejected adequate work because it differed from his preferred style. Perhaps the assignment's schedule made shared revision impractical. These possibilities cannot be resolved by counting final words attributed to each member. The instructor should examine the sequence and give the students a way to explain it.
A fair process might reveal that Leo made a necessary emergency correction but failed to involve teammates afterward. Another member may have missed an agreed deadline. The collaboration judgments can reflect distinct behaviors rather than assigning one person all blame or all credit. The team product remains a separate object of assessment.
This case shows why contribution percentages are often misleading. A student who writes most of the final text may have built on substantial work by others, or may have prevented others from doing the work. Volume is evidence about production, not a complete measure of responsibility, learning, or cooperative conduct.
Case four: a disruption changes what participation can mean
A fourth team experiences an unexpected disruption when one member becomes temporarily unavailable. The details are private, and the instructor should not require the student to disclose them to teammates beyond what is needed to coordinate the project. Formal accommodation and exceptional-circumstance procedures remain governed by the institution.
The immediate design question is how to revise responsibilities and learning opportunities. Can the student complete a different contribution within the approved arrangement? Does the team need a reduced scope or adjusted deadline? Which criteria remain applicable, and what evidence can fairly be collected? The answer should be worked out through the appropriate process rather than left to peer resentment.
The team should receive enough operational information to plan without becoming the judge of the student's personal circumstances. For example, they may need to know that a particular task has been reassigned and that the instructor has approved the revised arrangement. They do not need a private explanation that invites debate about whether the student deserves support.
Peer feedback must be interpreted in light of the approved arrangement. A student should not be penalized by peers for failing to attend a meeting they were formally excused from. Equally, the instructor still needs appropriate evidence for the learning claims attached to the grade. Support changes how evidence is obtained; it should not lead to an unexplained inference that learning occurred without any demonstration.
The case makes an important boundary visible. Teammates can report what happened in the work. They should not decide medical, disability, or other formal eligibility questions. The instructor must keep those responsibilities separate while ensuring that the remaining team has a feasible project.
Peer reports should ask for observations
A useful peer form asks about specific behavior within the assessment period. Did the person complete agreed work or communicate a difficulty early enough for the team to respond? Did they explain decisions that affected others? Did they engage with disagreement in a way that helped the project proceed? These questions still require interpretation, but they are more informative than “Was this person a good teammate?”
Ask for a brief example when a rating is unusually high or low. A numeric score without context can conceal different standards across students. One student may reserve the highest rating for exceptional work; another may give it to everyone to avoid conflict. The instructor should not assume that equal numbers have equal meaning.
The form should explain who will see the responses and how they will be used. Do not promise absolute anonymity if the situation or institutional process makes that impossible. Students need a truthful account of confidentiality, particularly when they report a conflict that could be recognizable from its details.
Peer reports are best treated as a source of evidence alongside individual work, project records, and instructor observations. They can reveal behavior the instructor did not see, but they can also reflect friendship, conflict, limited visibility, or coordinated ratings. The grading process needs a way to investigate material discrepancies without turning every project into a disciplinary hearing.
Do not calculate fairness by multiplying popularity
Some grading arrangements multiply the team score by a peer contribution factor. Such a method can be transparent mathematically while remaining difficult to justify educationally. A multiplier can affect credit for the quality of the product based on reports about behavior, combining two claims that may deserve separate treatment.
Consider a team product score of eighty and a peer factor of zero point eight. Multiplication produces sixty-four. The arithmetic is correct. The unresolved question is why the peer evidence justifies reducing the product score by sixteen points, and whether students understood and could fairly influence the behavior being assessed.
This does not establish that every peer-adjustment method is inappropriate. It means the instructor should explain the model's meaning, examine its sensitivity, and provide a route for reviewing disputed evidence. A small change in peer ratings may have a large effect when applied to the whole grade. That effect should be intentional rather than discovered after results are released.
In the fictional contract, collaboration is a separate weighted component. This makes its influence visible and lets feedback identify the relevant behavior. The design still requires judgment and can still be flawed, but it avoids pretending that a peer score is a precise estimate of each person's share of a collective product.
Collect enough process evidence, not everything
The instructor does not need a complete transcript of every team interaction. Excessive documentation can consume time intended for learning and may intrude on private communication. The course should choose a small set of evidence points that serve clear assessment purposes.
A team agreement establishes initial responsibilities and communication expectations. A midpoint note records important changes and unresolved dependencies. An individual reflection explains one contribution and one decision the student learned from. These artifacts can be concise. Their usefulness depends on whether they help assess the published criteria, not on their volume.
Version histories can clarify who edited a document, but they do not capture every kind of work. A student may conduct an analysis offline, facilitate a decision verbally, or help another person understand a difficult concept. Treating edit counts as a complete contribution measure would reward some activities while erasing others.
The Eberly Center's group-project design guidance emphasizes planning the conditions under which group work supports learning. In this example, process evidence is one such condition. It should make meaningful work assessable while leaving students room to collaborate naturally and protecting information that the instructor does not need.
Teach the behavior before grading it
Students may have little experience negotiating roles, giving constructive feedback, or revising a plan when someone falls behind. If those behaviors affect the grade, the course should provide instruction and practice rather than assume that students arrive with identical opportunities to learn them.
A short exercise can ask teams to respond to a fictional missed deadline. What information do they need? How can they revise the plan without making unsupported assumptions about the person? When should they involve the instructor? The exercise makes the expected process visible before a real conflict becomes emotionally charged.
Similarly, students can practice explaining a disagreement through evidence and consequences. “Your idea is bad” and “This option depends on data we do not have” are different contributions to a decision. The collaboration criterion should reward behavior that supports the task, not require a particular personality or constant agreement.
The instructor should also model how to ask for help. If students believe that reporting a problem will automatically damage the team's grade, they may conceal difficulties until the deadline. A clear escalation route can make early communication part of responsible work rather than a confession of failure.
Release grades that explain the three claims
When grades are returned, students see the product score, the individual-understanding score, and the collaboration score with focused feedback. This lets them distinguish a strong team artifact from a personal learning gap or a process issue. A single unexplained total would hide those differences.
Feedback should connect to evidence the student can understand. If a collaboration score is lower because of missed commitments, identify the relevant pattern and the published expectation, subject to confidentiality rules. Do not reproduce another student's private comments indiscriminately. The instructor owns the assessment judgment and should be able to explain it rather than merely say that peers supplied the number.
Students also need the established route for correcting factual errors or requesting review. A mistaken attribution, a missing submission, or an approved arrangement overlooked during marking can materially affect the result. The review should examine the evidence and criteria, not reward the most persistent complainant.
After the project, the instructor reviews whether the assessment produced the intended learning information. If individual tasks repeatedly reveal gaps hidden by strong products, the course may need more shared explanation during the project. If collaboration evidence is consistently ambiguous, the criteria or observation points may need revision. The next design should respond to those findings.
The project remains collective
Separating evidence does not require turning teammates into competitors. Students can still share ideas, specialize, revise one another's work, and take pride in a common result. The assessment simply avoids making claims that the shared artifact cannot support on its own.
A well-designed contract tells students what they are learning together and what each person must demonstrate. It recognizes that contribution can be quiet, that visible effort can still undermine collaboration, and that a polished product may coexist with uneven understanding. Those distinctions make the grade more useful because they preserve the complexity of the work instead of hiding it behind a single team score.