Rather than attempting to “AI-proof” assessments, consider designing assessments that emphasize process as well as product.  

For example:

• Have students submit drafts for review and provide feedback.
• Have students reflect on their process of creating a final product.
• Have students keep research logs.
• Have students design journal entries.

Students may use AI during project development but must submit a brief reflection describing:

• How AI was used
• What prompts they used to generate an output
• What outputs were accepted or rejected
• What independent decisions they made

Instead of testing recall, use AI to design tasks that require:

• Analysis
• Evaluation
• Decision-making
• Reflection
• Problem-solving

Traditional Question:  What are greenhouse gases?

Authentic Assessment:  A city plans to reduce carbon emissions by 40% within ten years. Evaluate three strategies and recommend the most effective approach, supported by evidence.

AI can quickly create versions for different learner needs.

Examples:

• Beginner
• Intermediate
• Advanced

Put the following prompt into Copilot:

“One learning outcome for a course is that students will be able to analyze the causes and consequences of climate change. The assignment to measure this learning outcome is as follows: 

A city plans to reduce carbon emissions by 40% within ten years. Evaluate three strategies and recommend the most effective approach, supported by evidence.”

Generate three versions of this assignment with increasing levels of complexity while maintaining the same learning outcome.  Create an assessment to evaluate this assignment.

Now Try This with an assignment for one of your courses!