Question Assumptions
What does the AI take for granted? Every answer rests on invisible premises. Make them visible before you act.
Sprint 3 · Module 7~12 Min
Critical Thinking.Use AI Consciously
Use AI consciously. Reflect before acting. Think strategically.
Discover
Self-Check
How Far Does Your Critical Awareness Reach?
Four levels – from blind trust to verified insight. Where do you stand?
1
Trusting
"The AI said it, so it must be true."
2
Doubting
"Sometimes I feel like something's not quite right."
3
Verifying
"I compare sources and test for consistency."
4
Mastery
"I use multiple AIs for cross-checks and form my own opinion."
"The unexamined AI output is not worth acting upon."
— PLU KI Academy
The 4 Steps
Critical Thinking with AI
A structured approach to evaluating AI output. Follow these steps and turn blind trust into informed judgment.
Check Sources
Where does the knowledge come from? AI blends training data without citations. Verify facts that matter.
Explore Counter-Positions
What speaks against it? The best way to test an idea is to try to destroy it. Ask the AI to play devil's advocate.
Test Consistency
Same answer with different prompts? Rephrase and re-ask. Inconsistent answers signal uncertainty or hallucination.
Toolbox
Your Critical Thinking Toolkit
These five tools transform you from a passive consumer into an active evaluator of AI output.
1
Fact-Check
Verify accuracy. Cross-check dates, names, statistics, and claims with reliable sources. AI can be confidently wrong.
2
Bias Detection
Identify distortions. AI reflects the biases in its training data. Watch for stereotypes, skewed framing, and missing perspectives.
3
Source Analysis
Evaluate origin. Ask where the information comes from. AI rarely cites sources — demand transparency for important decisions.
4
Logic Check
Find contradictions. Follow the argument step by step. Does A really lead to B? Are there gaps or fallacies?
5
Consistency Test
Self-correction. Rephrase your question and compare answers. Consistent responses are more likely to be reliable.
Interactive Exercise
Good or Bad? The Critical Prompt Test
Three scenarios. Choose the better approach and learn why it builds critical judgment.
1
You receive a detailed AI analysis about market trends. What do you do first?
2
The AI gives you a one-sided argument for a business decision. How do you respond?
3
You need to be sure about an important AI-generated fact. What's the best strategy?
Self-Test
Were you paying attention?
Three questions — test your critical thinking knowledge before moving on.
1
What is the first step in the critical thinking process with AI?
2
Which technique helps you discover bias in AI output?
3
Why is it important to explore counter-positions?
"Doubt is the beginning of wisdom, not the end of it."
— Aristotle
Live Example
From Analysis to Cross-Check
Four steps – from first assessment to verified insight. The cross-check across multiple AIs is the most important step.
Round 1: Initial Analysis
Human
Analyze the claim: "Remote work increases productivity." Give me a balanced view with pros and cons.
AI A (ChatGPT)
Balanced Analysis: Remote Work & Productivity
Pros: Fewer commutes mean more focused hours. Many workers report higher satisfaction and lower burnout. Access to global talent pools increases team capability.
Cons: Collaboration can suffer without spontaneous in-person interaction. Onboarding new employees is harder remotely. Some roles require physical presence and cannot be evaluated by the same metric.
Verdict: The effect depends heavily on role type, company culture, and measurement criteria.
Pros: Fewer commutes mean more focused hours. Many workers report higher satisfaction and lower burnout. Access to global talent pools increases team capability.
Cons: Collaboration can suffer without spontaneous in-person interaction. Onboarding new employees is harder remotely. Some roles require physical presence and cannot be evaluated by the same metric.
Verdict: The effect depends heavily on role type, company culture, and measurement criteria.
Round 2: Devil's Advocate
Human
Now play devil's advocate. Critique your own analysis. What did you miss? What assumptions did you make?
AI A (ChatGPT)
Self-Critique & Missing Angles:
Assumption 1: I assumed "productivity" is universally defined. In reality, output per hour, creative breakthroughs, and team cohesion are very different metrics — and remote work affects each differently.
Missed angle: The selection bias in studies. Companies that adopt remote work successfully may already have strong management cultures. The causality may be reversed.
Missed angle: Long-term career effects. Junior employees learn faster through osmosis in offices. Remote work may trade short-term output for long-term skill development gaps.
Weakness: I presented the verdict as balanced, but I did not quantify how often or for whom each scenario applies.
Assumption 1: I assumed "productivity" is universally defined. In reality, output per hour, creative breakthroughs, and team cohesion are very different metrics — and remote work affects each differently.
Missed angle: The selection bias in studies. Companies that adopt remote work successfully may already have strong management cultures. The causality may be reversed.
Missed angle: Long-term career effects. Junior employees learn faster through osmosis in offices. Remote work may trade short-term output for long-term skill development gaps.
Weakness: I presented the verdict as balanced, but I did not quantify how often or for whom each scenario applies.
Round 3: Consistency Check
Human
Rephrase: Does working from home make people more efficient? Compare with your previous answer.
AI A (ChatGPT)
Consistency Check:
Core claim holds: It depends. The nuanced verdict is consistent across both phrasings.
Refinement: Your rephrasing highlights "efficiency" rather than "productivity." Efficiency emphasizes output per hour — where remote work often wins. Productivity is broader and includes innovation, which may suffer remotely.
Self-correction: My first answer underweighted the measurement problem. Your rephrasing made it clearer that the question itself is ambiguous without defining the metric.
Conclusion: The answer is consistent but the second framing sharpened the focus on what actually matters — how you measure success.
Core claim holds: It depends. The nuanced verdict is consistent across both phrasings.
Refinement: Your rephrasing highlights "efficiency" rather than "productivity." Efficiency emphasizes output per hour — where remote work often wins. Productivity is broader and includes innovation, which may suffer remotely.
Self-correction: My first answer underweighted the measurement problem. Your rephrasing made it clearer that the question itself is ambiguous without defining the metric.
Conclusion: The answer is consistent but the second framing sharpened the focus on what actually matters — how you measure success.
Round 4: Cross-Check Across AIs ⭐ ESSENTIAL
Human
The same original question is now asked to a different AI. Compare the answers — where do they agree? Where do they contradict? What does AI B uncover that AI A missed?
AI B (Claude)
Comparative Analysis of AI Responses:
Agreement: Both AIs emphasize context-dependency. Both mention commute time and flexibility as pro arguments.
Contradiction: AI A was more optimistic about remote work; AI B emphasized the innovation gap and "Zoom fatigue" more strongly. AI B cited specific figures (15-25% productivity gains in studies), which AI A omitted.
What AI B uncovered:
• The "productivity illusion" — many remote workers work longer, not more efficiently
• The mentorship loss — junior employees learn slower without direct contact
• The asymmetric distribution — remote work favors certain personality types
➜ Bottom Line: No single AI delivers the whole truth. Cross-checking across different AIs (ChatGPT, Claude, Gemini, Perplexity) provides the most reliable path to a complete picture. Each AI has its own training data, blind spots, and strengths.
Agreement: Both AIs emphasize context-dependency. Both mention commute time and flexibility as pro arguments.
Contradiction: AI A was more optimistic about remote work; AI B emphasized the innovation gap and "Zoom fatigue" more strongly. AI B cited specific figures (15-25% productivity gains in studies), which AI A omitted.
What AI B uncovered:
• The "productivity illusion" — many remote workers work longer, not more efficiently
• The mentorship loss — junior employees learn slower without direct contact
• The asymmetric distribution — remote work favors certain personality types
➜ Bottom Line: No single AI delivers the whole truth. Cross-checking across different AIs (ChatGPT, Claude, Gemini, Perplexity) provides the most reliable path to a complete picture. Each AI has its own training data, blind spots, and strengths.
The Most Important Step: Cross-Check
An AI doesn't intentionally lie — but it can be wrong, produce hallucinations, or miss important aspects. The same prompt across ChatGPT, Claude, Gemini, and Perplexity yields four different perspectives. Where they all agree is likely correct. Where they contradict, the truth lies in the middle — or there is no clear answer at all. Cross-checking is not optional; it is mandatory for important decisions.
Your Result
What You Take With You
Question Everything
Assumptions, sources, and conclusions. The best AI users are the ones who never stop asking "How do you know?"
Seek Counter-Positions
The strongest ideas survive the strongest objections. Make devil's advocacy a habit, not an afterthought.
Test for Consistency
Rephrase and re-ask. Consistent answers are more reliable. Inconsistency is a red flag worth investigating.
Your Learning Path
Sprint 3: Master
Creativity
Master AI as a creative partner for brainstorming, writing, and ideation.
Critical Thinking
You are here. Learn to reflect, verify, and question AI outputs. The key to safe and effective AI use.
Glossary
The complete lexicon of AI terms and concepts. Your reference for everything you've learned.