AI Guides: AI for Feedback

🤖 AI Guides @ JCU

• AI Guides: Resources for Staff
• AI Guides: Redesigning Assessment to Integrate AI
• AI Guides: Designing Curriculum for Learning
• AI Guides: UDL and AI
• AI Guides: AI for Feedback
• AI Guides: Generative AI Tools at JCU
• Ethical practices for assessment task design
• Artificial Intelligence Policy

Using AI for feedback and marking: Ethical workflows, prompts, do/don't, and transparency language for Microsoft Copilot

2) Approved workflow (step-by-step)

Step 1: Read and grade the student work yourself

• Review the submission
• Apply the rubric, and determine the grade
• Make brief notes on strengths and areas for improvement.

Why: AI cannot make grading decisions. Your professional judgment comes first.

Step 2: Prepare a de-identified summary (if needed)

If you want AI help with feedback, create a brief summary of the work without any student identifiers.

Example: "Student addressed 2 of 3 criteria, provided limited evidence, conclusion was unclear."
Why: Protects student privacy; AI never sees the actual student work.

Step 3: Use Microsoft Copilot to draft feedback language

• Open Microsoft Copilot (logged in with JCU credentials).
• Input your de-identified notes and rubric criteria.
• Ask Copilot to help draft constructive feedback.

Why: Copilot can help articulate feedback clearly and save time on phrasing.

Step 4: Review, edit, and personalise the feedback

• Check the AI draft for accuracy, tone, and relevance
• Add specific examples from the student's work.
• Adjust language to match your teaching voice.

Why: Generic AI feedback is unhelpful. Your expertise and personalisation make it meaningful.

Step 5: Finalise and release feedback

• Once you are satisfied with the feedback, copy it into your LMS or grading system.
• Release to the student as you would any feedback.

Why: You take full responsibility for the feedback; it represents your professional judgment.

3) Effective prompts for Microsoft Copilot

Use these templates to generate helpful, constructive feedback drafts.

Prompt 1: Drafting balanced feedback

Use when: You want to articulate strengths and areas for improvement clearly.

"I am grading a [type of assessment] and need to provide constructive feedback. The student achieved [X grade] because they [brief summary of performance against criteria].

Please draft feedback that:

1. Acknowledges specific strengths related to [criterion 1, criterion 2]
2. Identifies areas for improvement in [criterion 3]
3. Provides 2-3 actionable next steps
4. Uses a constructive, encouraging tone appropriate for [year level] students

Keep the feedback to 150-200 words.“

Copy text above

Prompt 2: Explaining a specific gap

Use when: A student made a common error and you want to explain clearly.

"A student in [subject/unit] struggled with [specific concept or skill]. Their work shows [brief description of the issue]. Please draft 3-4 sentences explaining why [correct approach] is needed and how it differs from what they did. Use clear, student-friendly language.“

Copy text above

Prompt 3: Suggesting improvement strategies

Use when: You want to recommend concrete next steps.

"A student needs to improve their [skill/criterion] for the next assessment. Based on the rubric criterion '[paste criterion]', suggest 3 specific, actionable strategies they can practice. Make them practical for [year level] students in [discipline].“

Copy text above

Prompt 4: Rephrasing for clarity and tone

Use when: You've written feedback but want to improve clarity or soften tone.

"I wrote the following feedback for a student: '[paste your draft]'. Please rephrase this to be clearer and more constructive while maintaining the same key points. Use an encouraging tone appropriate for undergraduate students."

Copy text above

4) Do's and Don'ts (quick reference)

5) Quality checks before releasing feedback

Before sending AI-assisted feedback to students, verify:

• Accuracy: Does the feedback accurately reflect what the student submitted?
• Specificity: Have you added specific examples from their work (not generic statements)?
• Alignment: Does the feedback connect clearly to the rubric criteria?
• Tone: Is it constructive, respectful, and encouraging?
• Clarity: Will the student understand what to do differently next time?
• Completeness: Does it address all relevant rubric criteria?
• Appropriateness: Is it suitable for this student's level and context?
• No errors: Have you removed any AI hallucinations or irrelevant content?

6) Common pitfalls and how to avoid them

Pitfall: Generic, impersonal feedback
Solution: Always add 2-3 specific examples from the student's work. Replace "Your analysis could be stronger" with "Your analysis of the case study would be stronger if you had connected the financial data in Table 2 to your conclusion about market trends.“

Pitfall: AI suggests incorrect or irrelevant advice
Solution: Verify all technical content and discipline-specific advice. Remove anything that doesn't match your curriculum or the student's actual submission.

Pitfall: Feedback doesn't match the grade
Solution: Check that the tone and content of feedback align with the mark. A high distinction shouldn't have mostly negative feedback; a fail should clearly explain major gaps.

Pitfall: Over-reliance reduces marking quality
Solution: Use AI as a time-saver for phrasing, not as a replacement for your expertise. If you find yourself just copying AI output, you've gone too far.

Pitfall: Inconsistent feedback across students
Solution: Review a sample of your AI-assisted feedback to ensure consistency in depth, tone, and helpful guidance across different performance levels.

7) Transparency language for students

Include this or similar language in your course outline or assessment information:

Option 1: Course outline statement

"To provide timely and constructive feedback, I may use Microsoft Copilot to help draft feedback comments. I review, edit, and approve all feedback personally before it is released to you. All grading decisions are made by me using the assessment rubric. Your work is never uploaded to AI tools, and your privacy is protected.“

Copy text above

Option 2: Assessment information

"Feedback on this assessment may be assisted by AI tools (Microsoft Copilot) to help articulate comments clearly and efficiently. All feedback is reviewed and finalised by your marker to ensure it is accurate, relevant, and helpful. Grading is conducted entirely by academic staff using the rubric.“

Copy text above

Option 3: Brief note in feedback comments

"This feedback was drafted with AI assistance and reviewed by me to ensure accuracy and relevance to your work."

Copy text above

8) Microsoft Copilot setup (5-minute checklist)

1. Access Copilot through your JCU Microsoft 365 account (do NOT use the free consumer version)
2. Verify you are logged in with your institutional credentials (check the top-right of the Copilot interface)
3. Familiarise yourself with Copilot's conversation history and settings
4. Test with a sample prompt before using for real student feedback
5. Confirm that your institution's data protection policies cover Copilot use (check with IT or EDQS if unsure)

9) Time-saving tips (efficiency without compromising quality)

• Create a template for your Copilot prompts that includes your rubric criteria and assessment type—saves retyping.

• Use Copilot to generate feedback for the most common issues first (e.g., "lack of evidence," "unclear thesis")—then personalise.

• Keep a document of effective prompts that worked well for your assessments—reuse and refine over time.

• Batch similar student issues together when requesting feedback drafts (e.g., "5 students had weak conclusions").

• Use Copilot most for mid-range work, where feedback is extensive—high/low grades often need more personalised comments anyway.