Creating AI-Proof Assessments

The Shift from Detection to Design

Rather than fighting a losing battle with AI detection tools (covered in Module 4), educators are increasingly focusing on assessment design that either:

1. Makes AI less useful (process-based, personal, real-time)
2. Integrates AI transparently (AI allowed with disclosure)
3. Assesses what AI can't do (application, synthesis, performance)

This approach reduces stress on both students and teachers while maintaining academic integrity.

Why Detection Isn't Enough

The Detection Problem:

• AI detection tools have significant error rates
• False positives harm innocent students
• AI-written text can be easily modified to avoid detection
• Detection creates adversarial relationships
• Students who can't afford paid AI tools are disadvantaged

The Design Solution:

• Shift burden from catching to designing
• Create assignments where authenticity is built-in
• Focus on learning, not policing
• Reduce anxiety while maintaining rigor

Strategy 1: Process-Based Assessment

What It Is: Instead of only grading final products, assess the process students use to create them.

How to Implement:

Process Element	How to Assess
Brainstorming	Submitted mind maps, notes
Drafts	Required multiple drafts with changes
Research log	Documented source discovery
Revision history	Track changes enabled
Reflection	Written process reflection

Example: Essay Assignment

Traditional:

"Write a 5-paragraph essay on the causes of WWI. Due Friday."

Process-Based:

"Essay on WWI causes. Submit:

• Brainstorm (Tuesday)
• Thesis and outline (Wednesday)
• First draft (Thursday, in-class writing time)
• Peer feedback exchange (Friday)
• Final draft with track changes (Monday)
• Process reflection: What changed and why?"

Why It Works:

• AI can generate final products, not authentic processes
• Students develop actual writing skills
• You see thinking development over time
• AI use becomes obvious when process doesn't match product

Strategy 2: Personal and Local Content

What It Is: Connect assignments to students' personal experiences, local context, or current classroom discussions that AI doesn't have access to.

Examples:

Subject	Traditional	Personal/Local
English	Analyze symbolism in The Great Gatsby	Connect Gatsby's pursuit to a goal you've pursued
History	Explain the Industrial Revolution	Interview family member about work changes they've seen
Science	Explain photosynthesis	Document plants in school/home garden over 2 weeks
Math	Solve word problems	Create word problems using local businesses/data

Why It Works:

• AI doesn't know your students' lives
• Responses can't be easily found online
• Encourages authentic engagement
• Students can't claim AI wrote about their grandmother

Strategy 3: Real-Time Assessment

What It Is: Create conditions where AI assistance isn't practical or possible.

Methods:

In-Class Writing:

• Supervised writing sessions
• Prompts given at start of period
• Handwritten or school-device only

Oral Assessment:

• Individual conferences
• Verbal explanation of written work
• Follow-up questions

Live Performance:

• Presentations
• Demonstrations
• Lab work with explanation

Example: Modified Essay Assessment

"Part 1 (take-home): Research and outline your essay on climate change solutions. Part 2 (in-class, 45 min): Using only your outline, write your essay. Part 3 (conference): Be prepared to explain your thesis and respond to questions."

Why It Works:

• AI can't help in real-time supervised conditions
• Students must genuinely understand their work
• Follow-up questions reveal depth of understanding

Strategy 4: AI-Integrated Assignments

What It Is: Rather than prohibiting AI, require its use with transparency and critical evaluation.

How to Structure:

"For this assignment, you MUST use ChatGPT or similar AI. Submit:

1. Your initial prompt to the AI
2. The AI's response
3. Your analysis: What did the AI get right? Wrong? What's missing?
4. Your improved version with AI as starting point
5. Reflection: How did AI help and hinder your learning?"

Why It Works:

• Teaches critical AI literacy
• Removes incentive to hide AI use
• Requires higher-order thinking
• Prepares students for AI-integrated future

Sample AI-Integrated Assignments:

Subject	Assignment
English	Use AI to generate 3 thesis options. Evaluate each and develop your own.
History	Ask AI about [event]. Fact-check 5 claims using primary sources.
Science	Have AI explain [concept]. Identify 3 oversimplifications and correct them.
Math	Use AI to solve problem. Explain errors in its approach if any.

Strategy 5: Application and Performance

What It Is: Assess students on application of knowledge, not just demonstration of knowledge.

Examples:

Knowledge Demonstration	Application/Performance
Write about lab safety	Demonstrate safe lab procedure
Explain persuasive techniques	Create and deliver persuasive speech
Describe historical analysis	Analyze new primary source in class
Define programming concepts	Debug code in real-time

Why It Works:

• AI can provide information but can't perform
• Shows transfer of learning
• More authentic assessment of capability
• Harder to fake understanding

Redesign Framework: The REAL Test

Use this framework to evaluate if an assessment is AI-resistant:

Letter	Question	If No, Consider...
Real-time	Does any part happen in supervised conditions?	Add in-class component
Experiential	Does it connect to personal/local experience?	Personalize the prompt
Application	Does it require applying knowledge, not just stating it?	Add performance element
Layered	Does it have multiple parts showing process?	Add drafts, reflections

An assessment passing all four is highly AI-resistant.

Example REAL Redesign:

Original: "Write a biography of a Civil War general."

Redesigned:

• Real-time: Presentation of key findings to class
• Experiential: Choose general based on connection to your family history or local area
• Application: What lessons from this general apply to leadership today?
• Layered: Research log, outline, draft, presentation slides, reflection

Common Concerns and Responses

"This takes more time to create and grade." Yes, initially. But you'll spend less time worrying about detection and handling integrity cases. Templates and rubrics can be reused.

"Students need to practice traditional writing." They will—in the process-based approach. The supervised writing component ensures this.

"Some students will still cheat." Some always have. But these approaches make it harder and less appealing. Focus on the majority who want to learn.

"This won't prepare students for standardized tests." Standardized tests are given in supervised conditions. Your in-class components mirror this.

"What about students with accommodations?" Design accommodations into the process (extended time for in-class portions, alternative demonstration methods).

Building Your Assessment Library

Step 1: Audit Current Assignments

• Which are most vulnerable to AI?
• Which already have AI-resistant elements?

Step 2: Prioritize Redesign

• High-stakes assignments first
• Major assessments that determine grades

Step 3: Create Templates

• Process-based essay template
• AI-integrated project template
• Oral defense questions

Step 4: Share with Colleagues

• Collaborative redesign
• Shared rubrics and templates
• Department alignment

Key Takeaways

1. Shift from detection to design—build authenticity into assessments
2. Process-based assessment reveals thinking AI can't replicate
3. Personal and local content connects to experiences AI doesn't know
4. Real-time components remove opportunity for AI assistance
5. AI-integrated assignments teach critical literacy and remove hiding incentive
6. Application and performance assess what AI can't do
7. Use the REAL test: Real-time, Experiential, Application, Layered

:::