Could this be a simulation?

If a civilisation reaches the capacity to run simulations of its own history containing conscious inhabitants, and if a reasonable fraction choose to run many, then simulated experiences of a kind like ours vastly outnumber unsimulated ones.

Given that, and given no way to tell from inside which kind you are having, the reasonable estimate of being simulated is high.

So either the first condition fails, or the second fails, or the conclusion holds. The paper is careful that this is a disjunction and that it does not establish which branch is true. Its own suggestion, in “the dark forest of our current ignorance”, is to spread belief roughly evenly across the three.

The argument assumes that a simulation of a mind would have experiences, and the paper takes this as given. It is the unsettled question of machine consciousness: a 2023 survey of the science adopts the view behind it only as a working hypothesis, and names researchers who reject it.

It also relies on a principle for reasoning when you do not know which of many observers you are, and that principle has been contested. And it assumes that computers able to simulate minds at scale are physically possible, which the paper supports with estimates for designs nobody has built. A 2025 calculation by Franco Vazza finds that no universe with physics like ours could supply the energy, even for a low-resolution simulation of Earth alone; Bostrom replies that it assumes a far more complete simulation than the argument needs.

Whether any of these objections is decisive is itself disputed. Brian Weatherson argued in 2003 that none of five ways of reading the principle will carry the conclusion, and Bostrom replied in 2005. Jonathan Birch argued in 2013 that the argument trusts our evidence about the limits of computing while doubting our evidence about our own physical make-up. Each is a place where the argument requires something it does not establish.

One proposal for detecting a simulation looks for artefacts of discretisation. It assumes our universe is being run with the lattice methods physicists already use on their own simulations, and uses the high-energy cutoff of the cosmic ray spectrum to bound how coarse that lattice could be. What it says to look for is not the cutoff. It is a direction: the highest-energy cosmic rays arriving with their rotational symmetry broken along the axes of the grid underneath.

That tests a particular implementation rather than the hypothesis, and its authors note that the improved methods lattice physicists already use would mask much of what it looks for. A simulation built to keep its own physics consistent leaves nothing of that kind to find. Bostrom still rejects the charge that the hypothesis is unfalsifiable: evidence that makes either of the first two branches less likely makes the third more likely, which is evidence, though not an experiment.

If the simulation is consistent, everything inside it behaves exactly as physics describes, and every measurement anyone makes comes out the same.

That is the honest reason this question sits so deep on the map rather than in the middle of it: it is a question about what is underneath the physics, asked from inside, with no instrument that reaches.