Simulation Theory and the Admissibility of Ontological Assent

Simulation theory moves the ground, but it does not ground the move.

Simulation theory is often presented as though the possibility of a generated world were enough to license an ontological claim about ours. It is not — but the reason is narrower and harder than it first appears. The casual objection, that “generated” smuggles in unexamined work and that an output is not yet a world, points at something real, but it cannot carry the argument. Our world already exhibits the features relevant here: distinguishable states, transitions, boundaries, persistence, and measurement conditions. That regimehood is the datum, not a debt the simulationist owes. Whatever else is in dispute, it is not whether this world is a coherent regime. It is.

So the burden cannot be “show that generation produces worldhood.” The world’s worldhood is given. The burden is epistemic, and it is this: for the claim this world is generated to be more than a relocation of the ground, there must be some possible discriminant — some condition, observation, structural residue, or theoretical consequence that would stand differently depending on whether the generation relation obtained. A claim that licenses no such difference does no work as an empirical or computational claim about this world. It moves this world’s explanation to another world and adds nothing weighable in the move.

The serious form of the thesis makes this sharp rather than dulling it. The version worth engaging is not “a world could be generated, so perhaps ours was.” It is the computationalist premise that substrate-independent computation can constitute minds and furnish a full regime for them — the premise that simulation arguments need in order to bite. The argument here is not aimed at Bostrom-style conditional or probabilistic arguments as such. It is aimed at the stronger inflationary move from possibility, probability, or generatedness to ontological assent. Admissibility tests that move directly, and the test produces a clean dilemma.

If the generator is finite and implemented, then the thesis at least has a coherent place to look for discriminants: precision limits, modular artefacts, compression residues, update and boundary constraints, or representational regularities. This is the good horn: the claim has possible content. But possible content is all it has. It is then an ordinary empirical hypothesis about an unobserved cause, presently unconfirmed and perhaps permanently inaccessible to measurement. The correct stance toward such a hypothesis is to hold it live and unconfirmed, as we hold any unobserved-cause claim — not to treat it as an undeniable maxim about reality. Inaccessibility of the discriminant is not confirmation of the cause.

Nor does ordinary lawfulness, compressibility, mathematical describability, or regimehood by itself supply the missing trace. Those features may motivate inquiry, but they are not yet discriminants unless they favour generation over native law. A trace must discriminate. Otherwise the very coherence of the world is merely being reread as evidence for the hypothesis it was already assumed to support.

If the generator is infinite, self-grounding, timeless, or purely mathematical — requiring no implementation and leaving no remainder — then the claim has left ordinary computational simulation and entered another register: Platonism, modal realism, an ultimate-grounding theory, theology under another name. That is not automatically a defect, and it is not automatically false. If such a theory supplies consequences that would stand differently depending on whether its grounding relation obtained, it re-enters the discriminant test. If it supplies none, it may still be metaphysics, but it must be argued as metaphysics and cannot inherit the empirical or computational authority that made “simulation” sound like a claim about the furniture of the world rather than a claim about its ground.

Note what this dilemma does not rely on. It does not assert that a program must be run to generate a world. The mathematical Platonist denies exactly that, and an intuition about execution is no argument against them. The dilemma sorts by discriminant, not by execution — and the Platonist falls on the second horn not because they fail to run anything, but because, unless further consequences are supplied, they leave nothing that could have been otherwise. The sort cuts both ways: a finite, implemented generator that left nothing that could have been otherwise would fall on the second horn too, and for the same reason. Finitude is the usual mark of a discriminant, not a guarantee of one; it is the trace, not the type, that decides.

One prior is load-bearing and should be named rather than smuggled: under the methodological standard being used here, a hypothesis offering no possible discriminant earns no empirical or computational assent as an ontological claim about this world. This is not the claim that discriminant-free metaphysics is meaningless. It is not a neutral fact, and it is not pretending to be one. Someone who grants assent to discriminant-free metaphysics can resist the conclusion — but then the disagreement has been relocated, cleanly, to that prior and to the register in which the claim is being argued. That is itself progress. The argument does not pretend to stand on nothing.

This standard applies evenly. It is not a special weapon against simulation theory. The same discipline would withhold empirical or computational assent from any unobserved-cause posit that offers no possible discriminant: eternal-inflation multiverses, Everett branches, string landscapes, or any comparable theoretical posit. Such claims may remain live, fruitful, mathematically motivated, or inferentially attractive. They may guide research. They may organise a field. But under the methodological standard being used here, live theoretical utility is not yet ontological assent. The standard bites broadly, and it should.

This does not require global scepticism. Some commitments are not optional theoretical posits but presuppositions of inquiry: other minds, a stable external world, memory, and a usable past. They may not be secured by discriminants of the kind demanded from speculative ontologies, but inquiry cannot proceed without them. They function as transcendental or practical conditions of asking, testing, doubting, and correcting at all. Simulation theory is not like that. It is not presupposed by inquiry; it is an additional explanatory posit. The standard therefore applies to dispensable ontological additions, not to the indispensable background conditions without which no standard could be applied.

There is also a distinction between being admissibly claimed and earning assent in a given register. A claim is admissibly claimed when it declares its register and pays the burden of that register. A discriminant-bearing hypothesis may be admissible as empirical or computational inquiry. A discriminant-free grounding thesis may be admissible as metaphysics if it is openly defended as metaphysics. It may even earn metaphysical assent within that register. But it does not thereby earn empirical or computational assent, nor may it borrow the authority of those registers. Admissibility is the right to enter the discussion without equivocation; assent is always indexed to the standard under which the claim is being advanced.

What remains is modest, and I think correct. As an undeniable maxim about reality, simulation theory fails on both horns: empirically unconfirmed on the first, register-shifted on the second. As a live empirical hypothesis, it is legitimate and open. As metaphysics, it may be argued, but must be argued as metaphysics. As a modelling discipline — games, strategic scenarios, artificial environments, situation simulations — it is productive and ontologically light. The error is to borrow the authority of one reading for another.

The central objection is therefore not that simulation is impossible, nor that generation cannot make a world, nor that discriminant-free metaphysics is meaningless. It is that the thesis moves the ground without grounding the move. To relocate this world’s explanation to another world is not to explain it, unless the relocation makes a weighable difference or openly declares and defends a non-empirical grounding standard.

A generated-world claim must first be admissibly claimed: it must declare its register and pay the burden of that register. But ontological assent under the methodological standard used here is stronger than admissible assertion. It requires more than possibility, elegance, usefulness, or relocation. For empirical or computational assent, it minimally requires a possible discriminant; for metaphysical assent, it requires an openly defended metaphysical prior that does not borrow authority from computation, empirical hypothesis, or modelling success.