Beliefs in Progress
Treat every belief as provisional and update it toward greater accuracy
- Difficulty
- Advanced
- Time to result
- ~ongoing to results
- Steps
- 5
- Confidence
- 98%
Beliefs in Progress replaces the identity of being right with the goal of becoming more accurate. A belief is expressed with a confidence level rather than as an absolute fact, making new evidence an input to calibration instead of a personal threat. The practitioner identifies what evidence would strengthen or weaken the belief, actively seeks information, and adjusts confidence incrementally as evidence arrives. This avoids the false binary in which changing one’s mind requires moving from completely right to completely wrong. Framing decisions as bets reinforces the mechanism: resources are being committed on limited information about an uncertain future, so success depends on building the most accurate available model rather than affirming prior opinions.
Origin
Extracted from Masters in Business, where Annie Duke connects probabilistic poker thinking to beliefs, forecasts, investing, and everyday decisions.
Core principles
- 01Accuracy matters more than defending prior beliefs
- 02Every belief carries uncertainty
- 03New information should change confidence, not threaten identity
- 04Belief changes can be incremental rather than total reversals
- 05Explicit uncertainty creates information hunger
How to run it
- 1
Express confidence explicitly
Translate a categorical belief into a probability, range, or qualified statement that reflects uncertainty.
Pro tip Use language such as “I currently lean toward” or “I am about 70% confident.”
Watch out Do not use an arbitrary number to disguise weak reasoning as precision.
- 2
Define update triggers
Write down what evidence would raise or lower your confidence before seeking additional information.
Pro tip Include evidence that would challenge your preferred conclusion.
Watch out A belief that no conceivable evidence can change is not being treated as provisional.
- 3
Seek disconfirming information
Research competing explanations and invite informed people to present evidence against your current view.
Pro tip Ask people who disagree what they know that you may be missing.
Watch out Do not apply harsher standards only to evidence you dislike.
- 4
Calibrate the belief
Adjust confidence in proportion to the strength and relevance of the new evidence.
Pro tip Small evidence usually warrants a small update rather than a complete reversal.
Watch out Do not preserve the original number merely to appear consistent.
- 5
Act on the updated model
Choose the action with the best expected value under your current probabilities, while remaining ready to update again.
Pro tip Record the belief and decision together for later calibration review.
Watch out Acknowledging uncertainty does not mean avoiding a necessary decision.
In the wild
A trend-following skeptic changes “trend followers are idiots” into a qualified statement of skepticism. That language invites an enthusiast to explain the strategy, allowing both sides to test their assumptions and move closer to an accurate view.
→ Uncertainty opens an information exchange that categorical certainty would likely suppress.
A founder estimates a 65% chance that a feature will increase retention and specifies which experiment results would alter that estimate. After mixed evidence, the founder lowers confidence to 50% instead of declaring the feature either proven or useless.
→ The team updates proportionally and avoids defending an outdated forecast.
Common mistakes
Making beliefs part of identity
When personal worth depends on being right, contradictory evidence feels threatening and is more likely to be rejected.
Confusing uncertainty with ignorance
A confidence level can represent substantial knowledge while honestly acknowledging what remains unknown.
Demanding a full reversal
Evidence often warrants moving confidence modestly rather than switching instantly between absolute positions.
Is it for you?
Best for
It is best for people forecasting uncertain futures or making decisions from incomplete information.
Not ideal for
It is not a substitute for decisive action when immediate safety rules or settled operational procedures apply.
From the transcript
“That accuracy is what feels good.”
“your beliefs as in progress.”
“It makes me information hungry.”
From the episode
Annie Duke Discusses Decision-Making in Poker and Life
Annie Duke