Designing for It
Redesigning an assignment
to be AI-honest
Instead of ignoring or trying to work around AI, an AI-honest assessment acknowledges AI’s pressence and tries to account for it. You might think that your current assessments are either immune to the challenge AI brings or you might think that it’s too difficult to adjust your assessment to be AI-honest. This entry will help you identify vulnerabilities in your current assessments and help you rebuild it.
The concept
While we’ve spent years (maybe decades) crafting the perfect assessments for our courses’ learning goal, we have to face the fact that what may have worked before AI tools were widely available may not work the same today. The problem with some assessments is that while they may be aligned with student learning, they also reward effort that can be replicated with chatbots with minimal prompting from the student. When trying to make your assessments AI-honest, we ask how do we shift the assignment toward demonstration of process, situated judgment, and disciplinary identity.
“Write a 1000-word essay arguing for or against [position]. Cite at least five sources.”
“Bring a draft to class. In writing, explain one place where you changed your mind during revision and why.”
The prompt
Here is an assignment I currently use:
[Paste your assignment prompt here]
Please do three things:
1. Show me exactly how a student could use AI to complete this assignment with minimal original thinking. Be specific — include the prompts they might use.
2. Identify which learning outcomes this assignment is actually assessing versus which ones AI can now substitute for.
3. Suggest three concrete redesigns that preserve the core learning goal but require demonstrated process, situated judgment, or disciplinary knowledge that AI cannot replicate.
↑ Replace the italicized placeholders with your own details before submitting
Prompt variables
Anticipated results
The AI will attempt to complete your assignment and bring you face to face with how a student might make an AI-generated response. Keep in mind, it might be uncomfortable seeing how easy this process is, but it’s important to perform these reality checks as this technology continues to improve and become more available.
While we all strive for alignment between learning goals and assessment, there’s always room for improvement. AI is able to bring any gaps or misalignment to our immediate attention before assessments are assigned.
The AI is also capable of providing some suggestions to help you modify your assessment. These are just starting points, and often, AI is best used here to help brainstorm a handful of solutions until you, as the instructor, are able to identify a good solution or until you are inspired to make your own solution. Remember, AI isn’t here to fix your assessment, but it can help you recognize what changes you want to make.
Caveats
- LLMs are, by default, generic output generators. Any response you get from AI is improved with deeper context of you, your class, and your assessment. The more specific you are with your input, the more specific LLMs can be with their output.
- The output is a thinking partner, not a final draft. The output is less important than what you decide to do with it (whether you agree or disagree)
- Some of what the AI flags as “vulnerable” may be intentional on your part. You may decide the task is fine as-is.
- This prompt works best on assignments that have a clear written deliverable. For lab reports, design projects, or performances, you may need to adapt step 1 of the prompt.
Reflection questions
- Is there a specific thing you can look for that a student could only produce if they actually did the thinking?
- What do you think would be missing in an assignment written soley by AI?
- Are you assessing the product, the process, or the person? Does the assignment make that clear to students?
