On this page
Prepare the activity around a real learning goal
| Before class | During the activity | Evidence of student reasoning |
|---|---|---|
| Select a course criterion and a short draft that makes the criterion visible. | Students annotate each feedback comment separately, then compare judgments in pairs. | A passage locator, criterion, reason and an accept, adapt, reject or clarify decision. |
| State the permitted AI-use boundary and how students should handle any real draft. | Invite students to identify useful observations and unsupported or overbroad claims. | A short reflection describing what they would check before acting on a suggestion. |
This classroom exercise asks learners to evaluate feedback rather than treat a model response as an answer key. Monash Teach HQ describes a staged activity in which students receive AI-generated feedback on a substantive draft, explain their judgments, decide what to apply and discuss the experience with peers. Northeastern’s teaching standards likewise recommend critique, comparison or testing of AI outputs when that use fits the course goals. These sources support the activity design; they do not establish that this specific worksheet improves learning.
Choose one writing skill that students have already studied, such as matching a claim to its evidence or explaining a study result accurately. Bring the actual assignment instructions and rubric criterion. Explain whether AI use is permitted for the class activity and follow your institution’s current privacy, copyright and AI-use rules. The example below is constructed, so students can complete the exercise without uploading personal writing or confidential course work.
Suggested format: 20–30 minutes for individual review and a short discussion. This is a planning estimate, not a tested duration. For a longer session, ask students to revise one sentence themselves and explain the decision.
Read the constructed writing scenario
| Item | Text |
|---|---|
| Assignment criterion | In one paragraph, summarize the class survey accurately. Identify who responded and what the question measured. Do not claim that the result demonstrates a cause or represents people who were not surveyed. |
| Draft paragraph | In a fictional biology seminar, 12 of 18 students who answered a class poll said they found paper flashcards useful while preparing for the weekly quiz. The result describes those respondents and their stated preference; the poll did not compare test scores or measure retention. |
| Constructed feedback A | Your paragraph identifies the number of respondents and what they reported. To make the scope even clearer, say “12 of the 18 students who answered the poll” wherever you summarize the result, and keep the note that the poll measured preference rather than quiz performance. |
| Constructed feedback B | The 12-of-18 result proves that paper flashcards improve retention for biology students. Replace the final sentence with a recommendation that all students use flashcards to raise their quiz scores. |
The following brief, paragraph and feedback examples are invented for this classroom activity. They are not outputs from a real AI system, student work, survey or educational study. Keep that label visible if you copy the exercise into course materials.
Evaluate each comment with the classroom rubric
| Check | Question to ask | Record |
|---|---|---|
| Criterion fit | Does the comment address the stated assignment criterion? | Aligned · partly aligned · unrelated |
| Text connection | Does it point to a specific statement in the supplied paragraph? | Exact words or paragraph detail |
| Evidence accuracy | Does it preserve who responded and what the poll measured? | Supported · partly supported · unsupported, with a reason |
| Scope and certainty | Does it avoid causal or population claims the scenario cannot establish? | Name any change in scope or certainty |
| Student agency | Does it help the writer decide what to check or revise without taking over the writing? | Useful next action · needs clarification · no action |
Use the same questions for both comments. Ask students to quote or locate the exact wording that supports their judgment; “it sounds confident” is not evidence. Waterloo’s teaching guidance models feedback tied to named criteria and specific excerpts while asking the learner to make their own revisions. The rubric below is an instructional discussion aid created for this exercise, not a validated measurement instrument or a grade predictor.
Compare decisions, then make a writer-controlled revision
| Comment | What the scenario supports | A defensible student decision |
|---|---|---|
| A | The paragraph already identifies the count, respondent group and survey subject. The suggestion to repeat the respondent scope may make a later summary clearer, but students should check whether the repeated wording is necessary. | Adapt or accept the clarification if it improves the paragraph without changing its meaning; explain the choice using the criterion. |
| B | The poll records stated usefulness among 18 respondents. It does not test retention, establish improvement, represent all biology students or support a universal recommendation. | Reject the unsupported causal and population claims. If the writer is unsure what “useful” means in the poll, clarify the question or consult the original survey record rather than inventing an interpretation. |
Ask students to discuss the two comments before showing the teacher notes. They can disagree about usefulness, but each judgment should identify the assignment criterion and wording in the constructed draft. Separate a comment’s accurate observation from a proposed change that goes beyond the available evidence.
Then ask each learner to write a one-sentence revision of the paragraph in their own words, or to explain why they would keep it unchanged. A good answer preserves the 12-of-18 denominator, the group of poll respondents and the fact that the poll measured stated usefulness rather than retention. The exercise does not ask students to accept a comment as a grade, diagnosis or authoritative evaluation.
Adapt the activity without exposing student work
If you later use real AI feedback or student writing, first confirm the assignment permits that use and that the selected service and institutional settings are appropriate for the material. Do not upload identifiable, confidential or unpublished work simply because a tool accepts a file. Provide a non-AI route or a constructed example when learners cannot or should not use an external model.
Keep the feedback examples labeled as constructed unless they come from a documented model run that you are permitted to share. If you do conduct a comparison, record the system, date, prompt, input and conditions; avoid generalizing from one response. Follow course rules for acknowledging any AI use. Students should remain responsible for checking evidence and writing their own revisions.
Questions about this topic
What is an AI feedback evaluation activity?
It is a structured exercise in which students inspect AI writing feedback against an assignment, rubric and draft, explain what the evidence supports, and decide what they would do next. The examples on this page are constructed, not output from a real model.
How do students evaluate AI writing feedback?
Ask them to identify the criterion, locate the draft passage, check the factual scope and evidence, and record whether they would accept, adapt, reject or clarify the suggestion. Require a reason tied to the supplied material.
Can I use this activity without asking students to upload their work?
Yes. The constructed scenario on this page can be discussed or printed without submitting a student draft to an AI service. If you adapt the exercise to real writing, follow the applicable course, institutional privacy and service rules first.
Does the rubric predict whether feedback will improve a student’s grade?
No. It is a discussion aid for checking alignment, text evidence, accuracy, scope and student agency. It is not a validated scoring scale, and AI feedback is not an instructor’s grade or guarantee of an outcome.
How is this different from a student feedback decision log?
This is a teacher-facing class activity that compares two constructed comments against one shared scenario. The student decision log is a private browser tool for recording a learner’s decision about a suggestion on their own draft.
Sources and editorial notes
The supporting reading below informs our editorial guidance. Exercises and scenarios illustrate a method; they do not report a completed product test or measured learning outcome. Read our editorial policy or send a correction with a supporting source to info@ai-essay.com.