Category: epistemology
Type: argument
Associated thinker(s): Nick Bostrom
Original publication or approximate period: Bostrom, “Are You Living in a Computer Simulation?” (Philosophical Quarterly, 2003)
Difficulty: Intermediate
Themes: simulation, probability, skepticism
Related experiments: Brain in a Vat; Descartes’s Evil Demon; Russell’s Five-Minute World
Suppose future civilizations can run many detailed simulations of their ancestors, and that the simulated people have conscious experiences. In his 2003 paper, Nick Bostrom argues that at least one of these three claims is true:
The first two claims concern fractions of civilizations; the third concerns the fraction of observers with human-type experiences who are simulated. The formal argument is a disjunction: it does not establish that the third claim is the true one.
If simulated observers could vastly outnumber their biological counterparts, what does that imply—if anything—about the chance that our own experiences are simulated?
Bostrom takes a limited form of substrate independence as an assumption: a sufficiently detailed simulation of the relevant brain processes could support conscious experience. He also estimates that technologically mature civilizations could run very many simulations. From those premises he derives the three-way disjunction.
A separate step connects the proportion of simulated observers to our own credence. Bostrom’s “bland indifference principle” applies only if we lack relevant information that distinguishes simulated observers from biological ones. So the simulated-observer fraction is not an assumed premise; it is one possible result of the argument, and the first-person inference has its own condition.
Accepting the disjunction leaves open which branch is true. Bostrom argues that if the third branch is true, our credence that we are simulated should be close to one; in his conclusion, he says that under current ignorance it seems sensible to distribute credence roughly evenly among the three branches.
The first-person inference is debated. Brian Weatherson challenges how Bostrom’s indifference principle should be interpreted; Bostrom replies that the argument establishes the disjunction, not that the simulation branch is more likely than the others. Joe Carlsmith’s later analysis explores the conditions an observer class must satisfy before a simulation argument can support a first-person conclusion. These are live philosophical questions, not a settled verdict.
Bostrom does not conclude that we probably are simulated. He establishes a disjunction and separately gives a conditional credence claim if the third branch holds. The first branch means that very few human-level technological civilizations reach posthumanity; it does not by itself imply that humanity will go extinct soon. The argument is not direct observational evidence that we are simulated: it combines assumptions about consciousness and future computing with anthropic reasoning. It also does not show that simulated lives are unreal or meaningless.
Bostrom briefly considers selective simulations, including simulations of only one person, as extensions of the main argument. Later analyses ask which observer classes can support the reasoning. If you use a different class from Bostrom’s “people with human-type experiences,” state it explicitly; those extensions are not additional horns in the original trilemma.
Bostrom’s author-hosted PDF was retrieved and its relevant passages in Sections II–VII were checked directly, including the formal disjunction (Section IV), the “bland indifference principle” (Section V), and the conclusion (Section VII). The paper treats substrate independence as an assumption; it does not prove that simulated processes are conscious. The first-person credence step is conditional, and the paper’s conclusion does not say that the simulation branch is more likely than the other two. Weatherson’s published challenge is included through its abstract and Bostrom’s full-text reply; Carlsmith’s paper is a later analysis, not a consensus statement.
Research check: 2026-11-01. This is an original educational explanation, not a substitute for the primary texts or expert review. Additional comparisons are teaching prompts, not established historical variations.