Loading...
September 2, 2024 • 2hr 32min
Improve your decision-making, frameworks for learning, backcasting, and more | Annie Duke (#60 rebroadcast)
The Peter Attia Drive
In this episode, Peter Attia interviews Annie Duke, a former professional poker player and author on decision making. They discuss how poker serves as an excellent model system for decision making in the real world, where we often have to make choices with incomplete information and uncertain outcomes. Annie explains key concepts from her book "Thinking in Bets" and provides frameworks for improving decision making across various domains.
Improve your decision-making, frameworks for learning, backcasting, and more | Annie Duke (#60 rebroadcast)
0:00
-0:00
Key Takeaways
- Poker is a better model for real-world decision making than chess, as it involves incomplete information and luck in addition to skill
- Humans are not naturally wired to think probabilistically, but we can improve this skill with practice and awareness
- The "resulting" heuristic - judging decisions solely based on outcomes - is flawed and leads to poor decision making
- We should examine both wins and losses through the lens of luck and skill to truly improve decision making
- Backcasting - working backwards from a desired future outcome - is a powerful strategic planning tool
- Four levels of decision analysis:
- Only examine losses through luck
- Examine losses through luck and skill
- Examine wins and losses through luck and skill
- Deeply examine even good outcomes for potential improvements
- We should strive to make decisions quickly in most cases, but take time on important decisions using probabilistic thinking
- Understanding this decision-making framework allows you to "go fast on the straightaways and slow on the curves" in life
Introduction
In this episode, Peter Attia interviews Annie Duke, a former professional poker player and author on decision making. They discuss how poker serves as an excellent model system for decision making in the real world, where we often have to make choices with incomplete information and uncertain outcomes. Annie explains key concepts from her book "Thinking in Bets" and provides frameworks for improving decision making across various domains.
Topics Discussed
Poker vs Chess as Models for Decision Making (6:45)
Annie explains why poker is a better model for real-world decision making than chess:
- Chess has perfect information - all players see the full board state
- Poker involves hidden information and probabilistic thinking
- Real life often requires decisions with incomplete information
- Poker also incorporates luck/chance, like many real situations
"Chess is missing this particular element, which decision making in general has, which has to do with incomplete information." - Annie Duke
Humans and Probabilistic Thinking (12:30)
They discuss why humans struggle with probabilistic thinking:
- We're not evolutionarily wired for it
- It takes practice and conscious effort
- Even those trained in math/statistics can struggle in real-world application
- We should aim to "shift our distribution" towards more probabilistic thinking
Variable Reinforcement in Poker (19:15)
Annie explains the psychological draw of poker:
- Variable ratio reinforcement schedule - rewards come at unpredictable intervals
- This is highly addictive, like slot machines
- Players can rationalize losses and pump up wins
- The unpredictability of wins creates a powerful dopamine response
Luck vs Skill in Poker and Other Domains (32:15)
They explore the interplay of luck and skill:
- In the short-term, luck plays a big role in poker outcomes
- Over the long-term, skill differences become apparent
- This applies to many domains - investing, business, etc.
- The narrower the skill gap, the more luck influences short-term results
"If you took the Red Sox and you had them play a little league team, there'd be basically no influence of luck." - Annie Duke
The Decision Matrix and "Resulting" (1:10:30)
Annie introduces the decision matrix and the flawed "resulting" heuristic:
- 2x2 matrix: Good/Bad Decision vs Good/Bad Outcome
- "Resulting" - judging decision quality solely on outcome
- This ignores luck/chance and leads to poor conclusions
- Example: Criticism of Pete Carroll's Super Bowl play call
"Resulting is a heuristic simplifier. Works really well in chess. If I lose to you in chess, I played worse than you, works very poorly in poker, if I lose to you in poker, who knows?" - Annie Duke
Consequences of Avoiding Bad Outcomes (1:21:45)
They discuss how fear of bad outcomes impacts decision making:
- Leads to overly conservative choices
- Stifles innovation and risk-taking
- Creates a culture of "CYA" (cover your a*s) thinking
- We need to examine good outcomes as critically as bad ones
Poker as a Model System for Life (1:31:30)
Annie explains how poker serves as a "model system" for decision making:
- Provides clear feedback on decisions
- Allows for examination of luck vs skill
- Forces probabilistic thinking
- Skills learned in poker can transfer to other domains
Decision-Making in Leadership (1:35:15)
They explore how leaders often make and encourage status-quo decisions:
- Fear of blame for bad outcomes
- Difficulty in justifying innovative choices
- Example: Bill Belichick's decision-making evolution
- Leaders need to create cultures that reward good processes, not just good outcomes
Learning from Successes and Non-Events (1:39:30)
Annie highlights the importance of examining successes and "non-events":
- We often only deeply analyze failures
- Successes and non-events (like Y2K) offer valuable lessons
- Example: Analyzing D-Day invasion decision
- We should strive to learn from all outcomes, not just negative ones
Becoming a Good Decision Maker (1:43:00)
Annie outlines the first steps to improving decision making:
- Recognize our tendency to rationalize outcomes
- Actively seek out contradictory information
- Examine both wins and losses critically
- Develop a habit of thinking probabilistically
Elite vs Average Decision Makers (1:49:45)
They discuss what separates elite decision makers:
- Willingness to deeply examine all outcomes
- Ability to separate process from outcome
- Constant seeking of improvement, even after successes
- Embracing uncertainty and probabilistic thinking
Framework for Learning and Levels of Thought (1:52:15)
Annie presents a framework for learning and decision analysis:
- Four levels of thought in examining outcomes
- Importance of analyzing wins, not just losses
- Overcoming ego/identity protection in analysis
- Striving for "level 4" thinking - deeply examining even good outcomes
Self-Deception and When to Apply Deep Analysis (2:00:30)
They explore the human capacity for self-deception:
- We naturally protect our ego/identity
- Early success can lead to long-term self-deception
- Importance of applying deep analysis to major decisions
- The longer the feedback loop, the more critical deep analysis becomes
Challenges of High-Level Thinking with Subtle Feedback (2:11:00)
Annie discusses the difficulty of high-level thinking in certain domains:
- Some skills (like swimming) have slower/gentler feedback
- This makes improvement more challenging
- Importance of creating artificial feedback mechanisms
- The "soft landing" problem - not realizing how much room for improvement exists
Backcasting and Pre-Mortems (2:13:30)
They explore the power of backcasting and pre-mortems:
- Backcasting: Working backwards from a desired future state
- Pre-mortem: Imagining future failure and working backwards
- These techniques reveal potential obstacles and luck factors
- Allows for better strategic planning and risk mitigation
Conclusion
Annie Duke provides a wealth of insights on improving decision making by embracing probabilistic thinking, critically examining all outcomes, and using tools like backcasting. She emphasizes that while this framework is powerful, it doesn't mean every decision requires extensive analysis. Understanding when to "go fast" and when to "go slow" in decision making is key. By adopting these mental models, we can make better choices in an uncertain world across various domains of life.