Teaching Students to Question AI, Not Just Use It
The A.U.D.I.T. Protocol transforms students from passive consumers of AI outputs into critical validators—the shift from "Sherlock Holmes" to "Columbo."
What is the A.U.D.I.T. Protocol?
A pedagogical framework for developing critical AI literacy in secondary education (grades 6-12). Instead of prohibiting AI use, A.U.D.I.T. teaches students to systematically evaluate AI-generated content.
A.U.D.I.T.
- Analyze
- Use Knowledge
- Doubt
- Interrogate Limits
- Test
Analyze
Clarity & Coherence
Does the response directly address the question? Are explanations logically structured? Students learn to identify vague or circular reasoning.
Use Knowledge
Accuracy & Precision
Cross-check AI responses against existing knowledge. Are facts verifiable? Do numbers make sense? Students activate their foundational understanding.
Doubt
Detect Hallucinations
AI confidently presents fabricated information as fact. Students learn to recognize warning signs: invented sources, impossible dates, contradictory statements.
Interrogate Limits
Depth & Breadth
Does the response oversimplify? What's missing? Students identify superficial answers and gaps in reasoning.
Test
Verify & Validate
Use alternative sources, check edge cases, test with examples. Students develop systematic verification habits.
Three Core Ideas
From "Sherlock" to "Columbo"
Traditionally, students were investigators seeking solutions (Sherlock Holmes model). With AI providing instant answers, the role shifts: students become critical validators who question machine-generated responses (Columbo model)—"Just one more thing..."
Teacher as Orchestrator
The teacher's role evolves from knowledge transmitter to orchestrator of critical thinking. Guide students in evaluating AI outputs, curate meaningful scenarios, and cultivate systematic skepticism.
Dialogue, Not Dictation
Students engage AI through iterative questioning, comparing outputs, and identifying limitations. This constructive dialogue develops metacognitive awareness about both AI capabilities and human judgment.
Five Classroom-Ready Sequences
Progressive activities designed for middle school (grades 6-8) and high school (grades 9-12), each targeting specific critical thinking skills.
Style & Patterns
Skill: Recognize AI writing patterns
Students analyze texts to identify AI-generated content through linguistic markers, repetitive structures, and stylistic patterns.
View Sequence →Facts & Verification
Skill: Fact-check AI outputs
The "trap-setting" approach: students craft questions designed to make AI fail (impossible dates, nonexistent theorems).
View Sequence →Reasoning & Logic
Skill: Evaluate mathematical reasoning
Students examine AI solutions for logical gaps, unjustified steps, and circular arguments.
View Sequence →Bias & Assumptions
Skill: Identify embedded biases
Through carefully paired prompts (same context, different names), students discover how AI responses reflect societal stereotypes.
View Sequence →RED TEAMING - Security Awareness
Skill: Understand AI vulnerabilities
Advanced sequence exploring manipulation techniques used by security experts to expose AI weaknesses. Ethical framing essential.
View Sequence →Ready to implement? All sequences include student worksheets, teacher guides, and assessment rubrics.
Access Full ResourcesWhy Critical AI Literacy Matters
📊 Prohibition Doesn't Work
Research shows 70% of teachers prohibit AI tools, yet students use them anyway—in secret. The A.U.D.I.T. approach acknowledges reality and teaches responsible use.
🧠 Foundational Knowledge Remains Essential
You cannot critically evaluate what you don't understand. A.U.D.I.T. reinforces the necessity of mathematical foundations.
🎓 Future-Ready Skills
Students will work in AI-saturated environments. Teaching them to be critical validators prepares them for careers requiring human judgment alongside machine capabilities.
⚖️ From Production to Validation
The pedagogical challenge shifts from "find the answer" to "evaluate this proposed solution." This aligns with higher-order thinking skills in Bloom's taxonomy.
Your Feedback Matters
These sequences are evolving. If you test them in class, your observations—what works, what doesn't, what you've adapted—will be invaluable to me.
Contact: philippe.dupeyrat@icloud.com
About This Framework
The A.U.D.I.T. Protocol was developed by Philippe Dupeyrat, Mathematics Education Inspector, France.
Originally created for the French education system, this methodology is designed to be adaptable to international contexts.
Complete pedagogical resources (detailed worksheets, teacher guides, assessment rubrics) are available in French at: philipped79.github.io/audit-ia