Peer evaluation drafter

Peer evaluation is the one moment in a group project where honesty is formally requested and almost never supplied. Everyone gives everyone fives, the teammate who vanished for three weeks gets "great collaborator," and the exercise becomes theatre. The incentives explain it: you will take another class with these people, and the social cost of a hard review is immediate while the benefit is diffuse.

Inflation costs the person being reviewed most of all. Someone whose contributions consistently arrive late and half-finished, and who is told every term that they are a great teammate, gets to hear it for the first time from a manager who can fire them. Business school is close to the last environment where that information is cheap to receive.

The whole difficulty collapses once you stop evaluating people and start describing events, so the form is built entirely out of that move. There is no overall-comments box anywhere in it - that is where the boilerplate lives.

1. What worked, specifically. Positive feedback suffers from vagueness too. Two fields: the thing you watched happen, and what it made possible. People are frequently unaware of which of their habits are valuable, especially the quiet ones like documenting assumptions or absorbing an ambiguous task without being asked.
2. What did not, in three parts. Name the observable thing rather than the inferred motive - you saw a missed deadline, you did not see "didn't care." Name the consequence, because feedback without a cost attached is easy to file away as someone's preference. Then say what good would have looked like: "flagging on Sunday that Tuesday would slip" is a specific alternative.
3. The proportionality question. Was this raised during the project? Feedback that appears for the first time in a final evaluation is an ambush, and it is the main reason peer review breeds resentment. If you did not raise it you can still write it, but write it acknowledging that - which is honest about your own part, and is the difference between feedback and a grievance.
4. Five checks on yourself. Recency, one impression spreading across everything, retaliation, whether you would say this to their face, and whether the scores you are about to give match the words you just wrote. These stay in this tab and are never part of the draft.

Under each field is a short list of rules-based checks: does this name something concrete and datable, does it describe behavior rather than deliver a verdict on the person, does it stop at what you saw rather than why you think they did it, does it name an effect on the team or the deliverable, is the praise backed by something specific. They are simple text heuristics running entirely in your browser - no AI call, no external service, and nothing is uploaded to be checked.

They are shown per field and pass-only. There is deliberately no total, no percentage, and no quality grade for your feedback: a score would invite writing to the checker instead of describing what actually happened, which is the failure mode this whole page is built against. A check that does not pass is a prompt to reread the line, not a verdict on it - sometimes the honest sentence is the one the heuristic flags.