Revising One Sentence: Feedback on a Claim That Says Too Much
Feedback is information a student can use to change an inference. A generated comment can be fast and fluent, so the instructor's job includes checking that it points at what this student actually wrote.
A student summarizes an invented classroom study in which one group received praise and scored higher than a group that did not, and writes: the study proved that praise works. A comment such as good effort, be careful with strong words tells the student little. A comment such as one comparison in one class can support a narrower sentence, so what could you say about these students that the design allows? points at a specific inference and suggests a move. Feedback earns its name when a student can act on it, so the useful test of a comment is what a student does next.
One sentence to revise
Asking for a revision of a single sentence keeps attention on the piece that needs work. The student rewrites proved that praise works into something like in this class, the group that received praise scored higher, which suggests praise may help but does not show it works generally. The instructor then sees whether the comment was used, and a later task with a different study gives a second look. None of this shows the feedback caused any improvement. A student may improve for other reasons, and finding out would take a comparison.
Generated feedback adds speed and a risk. A system may produce comments quickly; the relevant workload includes reading, correcting and delivering them. The risk is that the comments sound right while misreading the student. A fluent paragraph might praise the writing, which was not the issue, or object to a mistake the student did not make, or skip the sentence that mattered. Before any comment reaches a student, an instructor can check it against two things: the student's actual words and the criteria the course uses. A comment that fails either check is rewritten.
A check before sending
Four questions provide a starting point. Does the comment point to something the student wrote? Does it name the criterion it applies, such as whether a conclusion fits the design? Does it suggest a next move and not only a verdict? Is it accurate about the concept? Asking these of a sample of generated comments also shows an instructor where a tool tends to drift in this course. Whether the checking costs more than writing the comments directly is a judgment each instructor makes.
Student work is personal information. An institution may have rules about where it can be sent, and a course may have norms of its own. This guide offers general design thinking and not legal advice. A cautious default is to ask before placing student work in any outside system, and to use invented answers when testing a tool. Feedback also works best inside a relationship in which a student can ask what a comment means, so a short reply from the student about what changed and why can show whether the comment landed.
An original classroom example
Original, unevaluated feedback example. An invented student answer says the study proved that praise works, although one class compared one praised group with one unpraised group. A generated comment says good effort and recommends careful wording. The instructor checks it against the student's sentence and the criterion (conclusion fits the design), rewrites it to ask what the design allows the student to say about these students, and has the student revise that sentence before meeting a different study.
Questions for your course
- What would you need to see to say a student used a piece of feedback?
- How would you check a generated comment against the student's own sentence?
- What separates a verdict from a next move?
- What are your institution's norms about sending student work to outside systems?