Using gen AI to (re)design assessments

The advent of generative AI has, in some cases, motivated educators to reconsider assessment task design. This article offers ideas for using Copilot to help with assessment task design or redesign. The process involves:

• Starting with learning outcomes

• Prompting Copilot

• Evaluating output

• Reviewing and tweaking the assessment task

• Developing assessment criteria/a rubric

• Applying a checklist

Starting with learning outcomes

As with assessment task design in any context, the starting point should be the course learning outcomes: what students should know, value, and be able to do as a result of taking your course. Assessment tasks can then be designed to allow you and students to see the extent to which the learning outcomes have been achieved. Have your course learning outcomes at hand.

Prompting Copilot

When you’re ready to start prompting Copilot, using your McGill credentials, log in to the secure version of Copilot. Check for the green security shield in the upper right corner of the Copilot window. Then, copy and paste the prompt below into Copilot (the coloured highlights will not copy over). Tailor the prompt so that it aligns with your teaching context.

Here’s a structure¹ for how a prompt can be crafted to include context and guardrails:

Persona: Give it a specific user persona (e.g., instructor, student)

Context: Provide detailed context (include examples)

Specific objective: Provide an explicit objective (i.e., the goal/task you want to achieve)

Guardrails: Provide parameters for output (e.g., length of the output, format, limitations/constraints)

Here’s an example that illustrates the different features of the prompt:

I’m an instructor at McGill University. I teach [academic discipline or topic]. I teach a course called […] for [1st, 2nd, …] year students. I’ve attached the course outline for the course. It includes the learning outcomes. Enrollment in this course tends to be around 50 students. I care a lot about my students’ ability to transfer their learning to new contexts, so I would like the assessment task to go beyond rote memorization. Can you develop an assessment task that addresses the [first two learning outcomes]? Give me a maximum of two examples to start with. Ask me questions about my request, one at a time, to help with creating the most appropriate output.

Reminders:  

• Provide your course learning outcomes either in an attachment or in the prompt field.  

• Keep in mind that an assessment task can address one or multiple learning outcomes. Be sure to prompt Copilot accordingly. 

• The more information you provide about your teaching context, the more likely you are to receive relevant output. 

Evaluating output

Copilot generates suggestions, not final assessment designs. You should evaluate all output against your course context, learning outcomes, and disciplinary expectations. Consider these reflection questions:

• Which suggestions from Copilot might you wish to implement and why?

• Which suggestions from Copilot would you not want to implement and why?

Reviewing and tweaking the assessment task

You can keep the exchange with Copilot going until you have an assignment task design that suits your teaching context or one that you would like to continue developing on your own. Here are examples of follow-up prompts:

• For the same course and learning outcome(s), can you design an assessment task to mitigate students’ use of generative AI?

• For the same course and learning outcome(s), can you develop an assessment task that requires students to use AI for only part of the assignment?

Developing assessment criteria/a rubric

Copilot can be used to help you create rubrics for your assessment tasks. To make a rubric that most thoroughly reflects your expectations might take several iterations. Not sure what to ask Copilot to change? Read Rubrics: The basics to learn about types of rubrics and strategies to design them to promote learning.

Applying a checklist

After developing an assessment task, you can use the checklist below to systematically identify aspects of the design that may benefit from further refinement. The checklist, developed with assistance from the secure version of Copilot, offers a manageable set of prompts (that are neither prescriptive nor comprehensive) to help you with systematic refinement. The checklist items can also be used to generate follow-up prompts for Copilot to help you further develop or revise the assessment task.

1. Alignment with learning outcomes

​​☐​ Are the learning outcomes clearly defined and appropriate for the level of learning (e.g., undergraduate, graduate)?

​​☐​ Is the assessment task designed to measure the intended learning?

​​☐​ Is gen AI being used to support—not replace—the development of core skills?

2. Assessment format and variety

​​☐​ Is the format appropriate for the learning context (e.g., quiz, oral presentation, case study, portfolio)?

​​☐​ Have you considered varied formats (e.g., critiquing AI-generated content, comparing AI analyses vs. human analyses)?

​​☐​ Are you providing students with opportunities to demonstrate their learning through a variety of assessment task types over time?

3. Relevance and context

​​☐​ Can you explain how the task connects to real-world practice or research in your discipline?  

​​☐​ Does the use of gen AI reflect current and/or emerging practices in your discipline?  

​​☐​ Are students guided to critically evaluate AI-generated content for bias, accuracy, and ethical implications?  

4. Ethics and academic integrity  

​​☐​ Have you provided explicit expectations for ethical use of gen AI in the task (if gen AI is permitted)?

​​☐​ Have you considered how gen AI might affect professional standards or conventions in your discipline?

5. Cognitive load and accessibility  

​​☐​ Is the task appropriate for students’ stage of learning and familiarity with gen AI tools?  

​​☐​ Are supports provided for students who may be unfamiliar with gen AI technologies?  

​​☐​ Does the task balance technical complexity with pedagogical clarity?  

6. Feedback and reflective practice

​​☐​ Does the task allow for meaningful feedback (e.g., instructor, peer, AI-assisted)?

​​☐​ Is there an opportunity for students to reflect on their learning and the role of gen AI?

​​☐​ Can students articulate how gen AI influenced their thinking or decision-making?

 Read more: Using generative AI in teaching and learning (TAP KB)

¹ Inspired by Schell, J. (2026, Jan. 22). Working smarter with AI: Applying skills to strengthen learning through human-AI collaboration. [Webinar]. Harvard Business Impact.

Back to top

McLennan Library Building 3415 McTavish Street Suite MS-12 (ground level), Montreal, Quebec H3A 0C8 | mcgill.ca/tap