Vibe Teaching — Issue #11
Designing AI-Powered Assessments (When AI Can Do the Work)
Let’s address the elephant in the room.
AI can now:
• write essays
• solve problems
• generate code
• explain concepts
• create projects
So what does it mean to assess learning anymore?
If AI can produce the answer…
What are we actually measuring?
The Breakdown of Traditional Assessment
Most assessments were designed for a world where:
• students worked alone
• answers were hard to find
• effort = output
That world no longer exists.
Now:
Students can generate high-quality answers instantly.
Which means:
👉 Output is no longer proof of understanding
This is the biggest shift in modern education.
The New Goal of Assessment
Assessment is no longer about:
❌ “Can the student produce an answer?”
It must become:
✅ “Can the student think about, explain, and improve an answer?”
This changes everything.
The Shift: Product → Process
Traditional assessments focus on:
👉 Final results
AI-powered assessments must focus on:
👉 Thinking process
You’re not just evaluating what students submit.
You’re evaluating:
• how they got there
• what they questioned
• what they improved
• what they understood
Example: Old vs New Assessment
❌ Traditional
“Write an essay explaining the causes of the Cold War.”
AI can do this instantly.
✅ AI-Powered Assessment
Step 1
Use AI to generate an explanation.
Step 2
Identify weaknesses, bias, or missing perspectives.
Step 3
Compare with another source.
Step 4
Improve the explanation.
Step 5
Reflect on what changed and why.
Now you’re assessing:
• analysis
• evaluation
• synthesis
• reasoning
That’s real learning.
The 4 Types of AI-Resilient Assessment
1️⃣ Process-Based Assessment
Students show:
• drafts
• revisions
• reasoning
• decision-making
You assess the journey.
2️⃣ Oral Explanation
Students explain:
• their thinking
• their choices
• their understanding
Hard to fake.
Easy to assess understanding.
3️⃣ Application-Based Tasks
Students apply knowledge to:
• new scenarios
• real-world problems
• unfamiliar contexts
AI can help—but cannot replace thinking.
4️⃣ Reflection-Based Assessment
Students explain:
• what they learned
• what they struggled with
• how they improved
Reflection reveals depth.
Make Thinking Visible
This is the most important principle.
If you can’t see thinking…
You can’t assess learning.
Ways to do this:
• require step-by-step explanations
• include AI interaction logs
• compare drafts vs final work
• ask “why” questions
• use reflection prompts
A Powerful Assessment Strategy
Try this structure:
👉 Generate → Critique → Improve → Defend
Students must:
• create an answer (with AI)
• analyze it
• improve it
• defend their choices
This reveals true understanding.
Practical AI Prompts for Assessment Design
Prompt — Redesign an Assessment
“Convert this assessment into one that evaluates reasoning and critical thinking instead of just output: [paste task]”
Prompt — Add Reflection
“Create reflection questions that reveal a student’s understanding and thinking process.”
Prompt — Build a Rubric
“Create a rubric that evaluates analysis, reasoning, and improvement instead of just correctness.”
Prompt — Generate Application Tasks
“Create real-world scenarios where students must apply this concept.”
The Biggest Mistake to Avoid
Trying to “catch” AI use.
This leads to:
• frustration
• mistrust
• wasted effort
Instead:
👉 Design assessments where AI use is part of the process
The Bigger Shift
Assessment is evolving from:
👉 “What did you produce?”
To:
👉 “How did you think?”
This is a better measure of learning.
And AI is forcing us to make that shift.
Final Reflection
AI didn’t break assessment.
It revealed its weaknesses.
The future of assessment belongs to educators who can measure:
• thinking
• reasoning
• improvement
• understanding
Not just answers.
That is Vibe Teaching.
Coming Next Issue
The Future Classroom: AI as the Interface
What happens when AI is no longer a tool…
But the primary way students interact with learning itself.
If you teach:
How confident are you that your current assessments measure real understanding—not just output?
That question matters more than ever.