Giving Feedback on Student Work Without Softening the Real Issue
Most feedback gets softened on the way out. A real problem, a weak thesis, an unsupported claim, gets wrapped in enough praise that a student walks away remembering the compliment and missing the one note that would have actually improved the work.
Warning
Cushioning a real problem with praise doesn't make it land more kindly. It just makes it easier for the reader to skip past the one thing they actually needed to hear.
The kindest feedback is the clearest one, not the gentlest one.
Why softened feedback doesn't teach
A vague, cushioned note protects the student's feelings in the moment, but it doesn't give them anything specific to fix. "Good effort, just needs a bit more support" describes a feeling, not an action. The student who reads it has no idea what to actually change next time.
Before: read the work, write a few sentences of encouragement, mention the issue once, softly, in the middle.
After: give Claude the work and your actual grading criteria, ask for feedback that names the specific problem plainly, and keep the note count small enough that nothing gets lost.
A process for feedback that names the real issue
- 1
Give Claude the actual work and what you're grading for
Paste the piece as written, along with the specific skill or criteria you want feedback on. Feedback on "everything" tends to default to generic, surface-level notes.
- 2
Ask for a fixed, small number of notes
Request two or three specific points, not an open-ended list. A short list forces the most important issue to actually be included, instead of getting diluted among ten minor ones.
- 3
Ask each note to name what to do, not just what's wrong
A note that only says "unclear" doesn't teach. Ask for feedback that pairs the problem with a specific direction: what a stronger version would include.
- 4
Read it back and cut any hedging language
Check for phrases like "might consider" or "could potentially." If the issue is real, state it directly; hedging is often where the softening creeps back in during your own edit pass.
“
Here's an undergraduate's draft thesis statement and opening paragraph for a research paper on urban housing policy. Give feedback focused only on whether the thesis is arguable and specific, not grammar, and keep it to three specific, actionable notes.
”
Inside Claude Tutorial
Naming the real issue plainly, without over-softening it, is transferable.
Giving feedback that's specific and direct, instead of cushioned past the point of usefulness, applies well beyond grading. The app has a full lesson on it, with practice that carries over to any feedback you give someone whose work you want to see improve.
When the work is genuinely close to strong
Not every piece of feedback needs three notes. Sometimes the honest read is that the work has one real gap and is otherwise solid, and padding the feedback with minor notes to seem thorough just dilutes the one that matters.
Tip
Ask Claude directly whether this piece has one dominant issue or several smaller ones. Feedback structured around a false assumption, that there must be three separate problems, reads as less useful than an honest single note.
“
This draft is close to strong overall. Read it and tell me honestly: is there one real gap keeping it from being excellent, or several smaller issues? Give me feedback shaped to whichever is actually true.
”Continue reading
- Turning Quiz Results Into What to Reteach, Not Just Who to Flag: spotting a pattern across many students' work, not just one piece at a time.
- Adapting One Explanation for Both Struggling and Advanced Learners: what to do once feedback reveals two students need very different next steps.
- Getting Started with Claude: the underlying habits, context, iteration, specificity, this entire process is built on.