Implement GitOps workflows with ArgoCD and Flux for automated, declarative Kubernetes...
npx skills add defi-naly/skillbank --skill "thinking-fast-and-slow"
Install specific skill from multi-skill repository
# Description
Daniel Kahneman's cognitive psychology framework for understanding decision-making, biases, and judgment.
# SKILL.md
name: thinking-fast-and-slow
description: "Daniel Kahneman's cognitive psychology framework for understanding decision-making, biases, and judgment."
dimensions:
domain: [decision-making, psychology, risk-assessment]
phase: [diagnosis, evaluation, post-mortem, planning]
problem_type: [cognitive-bias, overconfidence, prediction, judgment-under-uncertainty]
contexts:
- situation: "making a prediction or forecast"
use_when: "need to calibrate confidence, check for overconfidence, consider base rates"
- situation: "evaluating a past decision"
use_when: "identifying which biases may have influenced the outcome"
- situation: "assessing risk"
use_when: "checking if intuition (System 1) is reliable or if deliberate analysis (System 2) is needed"
- situation: "someone is very confident"
use_when: "testing whether confidence is justified by evidence or just coherent storytelling"
- situation: "planning a project"
use_when: "applying reference class forecasting to avoid planning fallacy"
combines_with:
- think-again # for updating beliefs after identifying bias
- black-swan # for tail risk and prediction limits
- antifragile # for building systems that survive prediction failures
- skin-in-the-game # for evaluating whose predictions to trust
contrast_with:
- skill: think-again
distinction: "TF&S diagnoses HOW we think wrong; Think Again focuses on CHANGING beliefs and intellectual humility"
- skill: hidden-potential
distinction: "TF&S is about judgment/decisions; Hidden Potential is about growth and skill development"
Thinking, Fast and Slow Framework
The Two Systems
The mind operates through two distinct modes of thinking:
SYSTEM 1 (Fast) SYSTEM 2 (Slow)
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Automatic Effortful
Unconscious Conscious
Fast Slow
Parallel processing Serial processing
Associative Rule-based
Skilled/habitual Flexible
Effortless Controlled
Pattern recognition Logical reasoning
Always on Lazy, avoids work
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Key insight: System 1 generates impressions, intuitions, and feelings. System 2 endorses, rationalizes, or overrides them. Most of our judgments and choices originate in System 1.
WYSIATI: What You See Is All There Is
System 1 constructs the best possible story from available information. It doesn't flag missing informationโit works with what it has.
Consequences:
- Overconfidence: Coherent story = high confidence (regardless of quality of evidence)
- Framing effects: Same information, different presentation = different conclusions
- Base-rate neglect: Vivid specifics override statistical reality
- Jumping to conclusions: Fast, but prone to error
Defense: Ask "What would I need to know to change my mind?" and "What information am I missing?"
Cognitive Biases & Heuristics
Anchoring
The first number you encounter heavily influences subsequent estimates.
| Scenario | Anchor Effect |
|---|---|
| Negotiation | First offer sets the range |
| Pricing | Original price makes discount seem larger |
| Estimation | Random numbers affect "rational" estimates |
Defense: Consider the opposite. Generate your own anchor first. Ask "If this anchor didn't exist, what would I estimate?"
Availability Heuristic
Probability judgments based on how easily examples come to mind.
Leads to:
- Overweighting recent, vivid, or emotional events
- Underweighting statistical base rates
- Risk perception driven by media coverage, not actual frequency
Defense: Ask "Am I judging frequency, or just memorability?" Seek base rate data.
Representativeness Heuristic
Judging probability by similarity to stereotypes, ignoring base rates.
Example: "Steve is meek, tidy, needs order, and has a passion for detail. Is he a librarian or a farmer?"
Most say librarianโbut there are 20x more male farmers than librarians. The base rate dominates.
Defense: Always ask "What's the base rate?" before considering the specific case.
Substitution
When facing a hard question, System 1 substitutes an easier one.
| Target Question | Substituted Question |
|---|---|
| How happy are you with life? | What's my mood right now? |
| Should I invest in this stock? | Do I like this company? |
| How likely is this project to succeed? | Can I imagine it succeeding? |
Defense: Identify the actual question. Resist the easy substitute.
Prospect Theory
People don't evaluate outcomes absolutelyโthey evaluate gains and losses relative to a reference point.
VALUE
โ
โ Gains
โ โฑ
โ โฑ
โ โฑ (diminishing sensitivity)
โ โฑ
โโโโโโโโโโโโโโโโโโโผโโโโโโโโโโโโโโโโโโโบ OUTCOME
โฑโ
โฑ โ
โฑ โ
โฑ โ Losses
โฑ โ (steeper = loss aversion)
โ
Key Principles
1. Reference Dependence
Outcomes coded as gains/losses relative to reference point (status quo, expectation, aspiration).
2. Loss Aversion
Losses loom larger than equivalent gains. Ratio โ 2:1.
- Losing $100 hurts about 2x as much as gaining $100 feels good
- Explains risk aversion for gains, risk seeking for losses
- Explains endowment effect (owning something increases its value to you)
3. Diminishing Sensitivity
The difference between $100 and $200 feels larger than between $1,100 and $1,200.
Four-Fold Pattern of Risk Attitudes
| GAINS | LOSSES | |
|---|---|---|
| High probability | Risk AVERSE (take the sure gain) | Risk SEEKING (gamble to avoid sure loss) |
| Low probability | Risk SEEKING (lottery tickets) | Risk AVERSE (insurance) |
This explains why people buy both lottery tickets AND insurance.
Overconfidence
The most significant cognitive bias. Manifests in three forms:
1. Overestimation
Thinking you're better than you are.
- 90% of drivers think they're above average
- Entrepreneurs overestimate success probability
2. Overplacement
Thinking you're better than others.
- Easy tasks: overplacement common
- Hard tasks: underplacement common (everyone assumes they're bad)
3. Overprecision
Excessive certainty in accuracy of beliefs.
- 90% confidence intervals contain truth ~50% of the time
- Experts often no better than novices at calibration
Defense: Use reference class forecasting. Track your predictions. Widen confidence intervals.
Planning Fallacy
Systematic underestimation of time, costs, and risks; overestimation of benefits.
Why it persists:
- Focus on the specific case, not the reference class
- Anchoring on best-case scenarios
- WYSIATI: don't consider what could go wrong
Reference Class Forecasting (the cure):
1. Identify appropriate reference class
2. Obtain statistics of that class
3. Use as baseline
4. Adjust for specifics of current case (minimally!)
Example: Don't estimate "how long will MY kitchen renovation take?" Ask "How long do kitchen renovations typically take?" (Answer: almost always longer than estimated.)
The Halo Effect
Global evaluation (like/dislike) influences perception of specific traits.
"I like the CEO" โโโบ "Their strategy must be good"
"The product succeeded" โโโบ "The decision was wise" (even if it was lucky)
"They're attractive" โโโบ "They must be competent"
Defense: Evaluate traits independently. Collect judgments from multiple people before discussion.
Hindsight Bias
"I knew it all along" after learning the outcome.
Problems:
- Makes us overestimate predictability of events
- Prevents learning from surprises
- Unfair to decision-makers ("obvious" in hindsight)
Defense: Record predictions BEFORE outcomes. Conduct pre-mortems, not just post-mortems.
Regression to the Mean
Extreme performance is followed by less extreme performanceโnot because of causation, but statistical inevitability.
Misinterpretations:
- "Punishment works better than praise" (performance naturally regresses after extremes)
- "The Sports Illustrated cover jinx"
- "Treatment worked!" (patient was measured at their worst)
Defense: Expect regression. Don't over-explain it with causal stories.
Pre-Mortem Technique
Before starting a project, imagine it has failed. Ask: "What went wrong?"
Benefits:
- Overcomes groupthink
- Legitimizes doubt
- Surfaces risks that optimism suppresses
- Uses prospective hindsight
Application Framework
When making decisions:
1. Identify System 1 Signals
- Strong intuition or gut feeling?
- Immediate answer that "feels right"?
- Emotional response?
2. Engage System 2 Checks
- [ ] What's the base rate for this type of outcome?
- [ ] What information am I missing? (WYSIATI)
- [ ] Am I anchored on an irrelevant number?
- [ ] Am I substituting an easier question?
- [ ] What's the reference class forecast?
- [ ] What would a pre-mortem reveal?
3. Calibrate Confidence
- [ ] Would I bet money at these odds?
- [ ] Widen your confidence interval
- [ ] Remember: coherence โ truth
4. Protect Against Loss Aversion
- [ ] Am I avoiding a wise risk because losses loom large?
- [ ] Would I take this deal if framed differently?
- [ ] Evaluate gains and losses by their actual size, not their feeling
Quick Reference: Common Biases
| Bias | Definition | Trigger Question |
|---|---|---|
| Anchoring | First number influences estimates | "What's my anchor? Ignore it." |
| Availability | Easy to recall = seems common | "Memorable or actually frequent?" |
| Representativeness | Similarity over base rates | "What's the base rate?" |
| WYSIATI | Conclusions from incomplete data | "What's missing?" |
| Loss Aversion | Losses hurt 2x gains | "Am I overweighting potential loss?" |
| Overconfidence | Excessive certainty | "What's my track record?" |
| Planning Fallacy | Underestimate time/cost | "What's the reference class?" |
| Halo Effect | Global impression bleeds to specifics | "Am I conflating traits?" |
| Hindsight | "Knew it all along" | "Could I have predicted this before?" |
| Sunk Cost | Past investment influences future | "Would I choose this if starting fresh?" |
# Supported AI Coding Agents
This skill is compatible with the SKILL.md standard and works with all major AI coding agents:
Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.