Reference class

A reference class is the set of observers among whom you treat your own observation as one sample. ‘All humans who will ever live’ is one class. ‘All biological minds,’ ‘all intelligent beings,’ and ‘all moments experienced by any mind’ are others. The choice sounds like bookkeeping until a probability is divided by the size of the class. Then it becomes the engine of the result.

The Doomsday Argument is the clean example. Your birth rank is around 117 billion among Homo sapiens. If the class is all humans, that rank carries information about the eventual human total. If future digital minds count, one civilization might create trillions of observers and make your early biological position extremely unusual. If only observers with your background information count, the class shrinks again. The same evidence has not changed; the denominator has.

There is no universally accepted rule for choosing the class. Similarity helps, but similarity depends on the question: species may matter for birth ranks, chemistry for habitable universes, and cognitive architecture for Boltzmann brains. Choosing the class after seeing which answer it produces is circular. Choosing it before the calculation is principled but still contestable.

A good anthropic argument therefore names its reference class in the open and checks whether reasonable alternatives overturn the conclusion. The observer lottery embedded in the companion essay does exactly that: it keeps the observed rank fixed and lets the population move, exposing how much of the result came from the evidence and how much came from deciding who counts.