Detecting Cognitive Bias

🔄 Quick Recall: In the last lesson, you learned structured decision frameworks — weighted matrices, PCM analysis, and decision trees. Now let’s identify the biases that can corrupt those frameworks if you’re not watching for them.

The Bias Problem

Cognitive biases aren’t failures — they’re features of human cognition that usually help us process information quickly. The problem is that they operate automatically, even when careful analysis is needed. You can’t simply “decide not to be biased.” But you can create systems that make biases visible.

The AI Bias Audit

Use this prompt whenever you’re making a significant decision:

I'm about to make a decision and want to check for cognitive biases:

My decision: [what you're deciding]
My current leaning: [which option I'm favoring]
My reasoning: [why I prefer this option]

Check my thinking for these common biases:

1. CONFIRMATION BIAS: Am I only considering evidence that supports my preference? What evidence contradicts it?
2. ANCHORING: Is my thinking anchored to a specific number, first impression, or early piece of information?
3. SUNK COST: Am I influenced by past time, money, or effort I've already invested?
4. STATUS QUO BIAS: Am I favoring the current situation simply because it's familiar?
5. OVERCONFIDENCE: Am I more certain than the evidence warrants?
6. AVAILABILITY: Am I overweighting vivid or recent examples?
7. FRAMING: Would I decide differently if the same information were presented differently?

For each bias you detect, explain how it's affecting my reasoning and suggest a corrective question I should ask myself.

✅ Quick Check: Why tell AI your “current leaning” when checking for biases?

Because AI can’t detect bias without knowing your preference. If you’re leaning toward Option A, AI can check: Are you overweighting Option A’s pros? Are you dismissing Option B’s advantages? Are you anchored to something about Option A (the first offer, a friend’s recommendation)? Without knowing your leaning, AI can only describe biases in the abstract — not catch the specific ones affecting YOUR reasoning right now.

Devil’s Advocate Prompting

Force AI to argue against your preferred option:

I'm leaning toward [your preferred option] for this decision:

Decision: [describe it]
My reasons for preferring this option: [list your reasons]

Now argue AGAINST my preference as strongly as possible:
1. What am I missing or underweighting?
2. What's the strongest case for a different option?
3. What would have to be true for my preferred option to be the WRONG choice?
4. What are people who chose similarly and regretted it say about their mistake?
5. If I had to defend the opposite choice, what would I say?

Be genuinely challenging — don't give me a weak counterargument just to be balanced.

Pre-Mortem Analysis

Instead of asking “what could go wrong?” AFTER deciding, imagine failure BEFORE deciding:

I'm about to choose [option]. Help me run a pre-mortem:

Imagine it's one year from now and this decision turned out to be a DISASTER.

1. What went wrong? List the 5 most likely failure scenarios.
2. For each scenario:
   - How likely is this? (percentage estimate)
   - What early warning signs would I see?
   - What could I do NOW to prevent or prepare for it?
3. Which of these failure scenarios am I currently ignoring or dismissing?
4. What's the total probability that at least ONE of these scenarios happens?
5. Given this analysis, should I still proceed, modify my plan, or reconsider?

✅ Quick Check: Why is a pre-mortem more effective than simply listing risks?

Because framing matters for cognition. When you ask “what could go wrong?” your brain stays optimistic — it generates risks reluctantly and minimizes them. When you say “this failed — why?” your brain shifts into explanatory mode and generates failure scenarios more freely and vividly. Research by psychologist Gary Klein shows that pre-mortems increase the ability to identify reasons for failure by 30%. Same question, different frame, dramatically better results.

Bias-Specific Countermeasures

Against Sunk Cost Bias

I've invested [time/money/effort] in [project/relationship/career path] and I'm wondering if I should continue or stop.

Help me separate sunk costs from forward-looking analysis:
1. What have I already invested that I cannot recover? (This is irrelevant to the decision)
2. What will it cost to continue from HERE? (This IS relevant)
3. If I were starting fresh today with no prior investment, would I choose this path?
4. What's the opportunity cost of continuing? (What else could I do with that time/money/effort?)

Against Overconfidence

I'm [percentage]% confident in my decision to [choice].

Challenge my confidence level:
1. What specific evidence supports this confidence level?
2. What would I need to see to reduce my confidence to 50%?
3. How often are people this confident actually right? (base rate)
4. What am I assuming that I haven't verified?
5. Who disagrees with me, and what do they know that I might not?

Exercise: Audit Your Current Decision

Take the decision from Lesson 1:

1. Run the full bias audit prompt with your current leaning
2. Use the Devil’s Advocate prompt to argue against your preference
3. Run a pre-mortem on your preferred option
4. After these exercises, has your leaning changed? Why or why not?

Key Takeaways

• Cognitive biases operate automatically — you can’t “decide not to be biased,” but you can create systems that make biases visible
• Tell AI your current leaning when checking for bias — it can’t detect bias without knowing your preference
• Devil’s Advocate prompting forces genuine counterarguments against your preferred option — not the weak ones your brain naturally generates
• Pre-mortems (“imagine this failed — why?”) are 30% more effective at identifying risks than forward-looking risk lists
• Sunk cost analysis separates irreversible past investments from forward-looking decisions — “if starting fresh today, would I choose this?”
• Overconfidence is fought by asking “what evidence supports this confidence level?” and “what would change my mind?”

Up Next: In the next lesson, you’ll tackle risk and uncertainty — scenario planning, expected value analysis, and making decisions when you can’t know the outcome.