The Birth-Rank Bayesian Update
Use your observed rank to update between short and long futures
- Difficulty
- Expert
- Time to result
- ~months to results
- Steps
- 6
- Confidence
- 97%
Start with two hypotheses about a sequence's final size and assign priors using ordinary evidence. The urn analogy makes the mechanism clear: drawing ball eight is much more likely from an urn of ten balls than from an urn of one million. The Carter-Leslie doomsday argument maps total human population to urn size and one's birth rank to the observed ball. A relatively early birth rank is more likely if humanity ends comparatively soon than if quadrillions of humans eventually spread across space, producing a pessimistic Bayesian update. The difficult step is indexical epistemology: the method needs a defensible reason to treat oneself as if randomly sampled from a reference class of observers. The framework is therefore a structured update, not an uncontested forecast.
Origin
Nick Bostrom explains the Carter-Leslie doomsday argument through a numbered-ball urn analogy in this interview.
Core principles
- 01Begin with priors grounded in ordinary evidence
- 02Treat observer position as additional evidence
- 03Compare how likely the same rank is under each hypothesis
- 04Keep controversial sampling assumptions visible
How to run it
- 1
Define total-size hypotheses
Create at least two hypotheses for how large the completed sequence will be.
Pro tip Simplify to two hypotheses first to expose the structure.
- 2
Set empirical priors
Assign starting probabilities from non-indexical evidence such as known risks and growth prospects.
Watch out Do not smuggle the observed rank into the priors.
- 3
Observe your rank
Identify the observer's approximate position in the sequence.
Pro tip An order-of-magnitude rank may be sufficient.
- 4
Compare likelihoods
Estimate how surprising that rank would be under each total-size hypothesis.
Pro tip Use the numbered-ball urn as a sanity check.
- 5
Update the hypotheses
Apply Bayes' theorem so hypotheses under which the rank is less surprising gain probability.
Watch out The result is an update, not certainty.
- 6
Audit the reference class
Test whether treating the observer as a random sample from the chosen population is justified.
Pro tip Try alternative reference classes and compare sensitivity.
Watch out The conclusion can depend heavily on disputed indexical assumptions.
In the wild
A coin selects either an urn containing balls 1-10 or one containing balls 1-1,000,000. Before drawing, each urn has equal probability. Drawing ball eight is far more likely under the ten-ball hypothesis, so Bayes' theorem strongly updates toward the smaller urn.
→ A low observed rank makes the smaller total more probable.
Common mistakes
Ignoring the prior
Birth rank supplements ordinary risk estimates; it does not erase prior evidence.
Hiding the sampling premise
Reasoning as a random human from all humans who ever exist is substantive and contested, not a neutral default.
Is it for you?
Best for
It is best for advanced probabilistic reasoning where who, when, or where the observer is carries information.
Not ideal for
It is not ideal when there is no defensible reference class or sampling assumption.
From the transcript
“using base theorem that allows you to infer that it's now much more likely that the earn has only 10 balls than a million because…”
“you can observe your own birth rank your sequence amongst all humans who have ever been born.”
“you do a similar basin update and end up with a doomsday [snorts] argument conclusion which is that doomsoon hypothesis are much more probable than…”
From the episode
Nick Bostrom: How Entrepreneurs Can Win in an AI-Dominated World
Nick Bostrom