Why do some philosophers reject modal realism?

Question

To me, modal realism seems particularly parsimonious - it assumes that if our world can exist, then any world that could exist does exist in "total reality", if only because to say "only our world exists" would require additional boundary conditions that somehow exclude all other worlds as being able to exist. These are extra elements in a theory that we cannot say exist, whereas I know at least one world exists.

I know others will say "you are positing all these other universes that we don't know exist". True - but we get a lot of nice explanatory power from that assumption and it doesn't presuppose entities that we don't know exist (worlds/universes) - only that we cannot think of any mechanism by which other worlds cannot exist.

For example, with modal realism, we can provide an objective and deterministic basis for probability/randomness.

Anyway, was wondering what the main arguments are against modal realism given all its strengths.

Answer 1

The SEP has a bit about problems for modal-world concretism à la Lewis; in the immediately preceding text, they say:

Perhaps the biggest — if not the most philosophically sophisticated — challenge to Lewis's theory is “the incredulous stare”, i.e., less colorfully put, the fact that its ontology is wildly at variance with common sense. Lewis faces this objection head on: His theory of worlds, he acknowledges, “does disagree, to an extreme extent, with firm common sense opinion about what there is” (1986, 133). However, Lewis argues that no other theory explains so much so economically. With worlds in one's philosophical toolkit, one is able to provide elegant explanations of a wide variety of metaphysical, semantical, and intentional phenomena. As high as the intuitive cost is, Lewis (135) concludes, the existence of worlds “ought to be accepted as true. The theoretical benefits are worth it.”

Indeed, as the SEP entry on hyperintensionality observes:

Quine and Davidson’s idea that serious philosophy should use only extensional notions faded in the latter half of the twentieth century, which witnessed an intensional revolution: a collective effort to analyze notions which are fundamental for our understanding of the world and of ourselves, like belief, information, knowledge, meaning, content, essence, explanation, via a single theoretical system: employing possible worlds and constructions out of them. Standard possible worlds semantics (SPWS) found applications in logic, linguistics, game theory, artificial intelligence: a success story of philosophy.

Yet so I can think of my own reason for doubt (inspired in part, though, by the thinking of a major deviant set theorist, Thomas Forster):

• It's hard enough to say that the "whole world" as we know it, has a beginning, has its own definite/absolute boundary conditions, etc. And if there was some sort of beginning, if not of absolutely everything, of something very important even so, then it seems hard to explain how this beginning transpired (i.e. witness the problematique of the initial and accelerated cosmological expansion). If it's hard for worlds to exist, so to speak, why would we expect very many worlds to percolate into stability out of the meta-quantum ether of modality? For if there are all logically possible worlds, then since there are temporal-logically possible worlds with beginnings, yet there are as many without beginnings, and so as many as have stronger, more stable boundaries. So there will be a great many worlds that also come to an end, for ∃x¬∃y(x≺y) is a permissible factoid in temporal meta-logic; or, then, the issue is not the mere existence, but the stable, orderly existence, of so many worlds. (E.g. Lewis countenances quantifying over 2^2ℵ₀-many concrete possible worlds, as in bijection with all sorts of permutations of continuous manifolds or something; but as many a topologist and physicist can well tell you, there is very much that is mathematically, or even physically, possible, that yet would spell the death of any known kind of sentient being who was subjected to the processes involved.)

Answer 2

IS it more parsimonious? A good Bayesian test of parsimony is minimum description length (MDL). We have our observations - taken entirely from the universe we right now inhabit - and we wish to find a computer program of minimum length that outputs them exactly.

Can this computer program be made shorter if it is written with the assumption that other "possible worlds" exist?

First of all, all that matters is the length in bits required to output observations in the current world. We don't care about any other worlds except to the extent that they are useful assumptions to output the observations.

Perhaps the computer program could generate an array of worlds according to some universal rules, and then select the array entry that matches the observations. This probably wouldn't be all possible worlds, though; just ones following certain rules that are convenient to generate the observations. For example, the program could generate worlds from the many-worlds interpretation of quantum physics, and then select our own "branch." That would be reasonable. All these worlds would share the same laws of physics, though, so I'm not sure this would meet your criterion of all possible worlds.

would require additional boundary conditions that somehow exclude all other worlds as being able to exist.

You need to add those boundary conditions anyway, because your MDL program needs to exclude all the other worlds and only select the one that generates the exact observations in question.

For example, with modal realism, we can provide an objective and deterministic basis for probability/randomness.

No you can't. Are you imagining that each possible world would have an associated number, indicating the probability of that world? But these numbers would not match the Bayesian credences of any particular agent in any particular world. Probability is fundamentally a tool that agents can use to reason about effective actions, and probability depends on the agent. It quantifies what the agent doesn't know. "Objective" probabilities are problematic.

But what I'd say is the biggest problem with modal realism (and modal logic in general) is the question of exactly which worlds are "possible." Is any world I can describe possible? Is the Harry Potter universe possible? Where are the limits? It's reasonable (as described above with MDL and MWI) to suppose a multiverse with certain laws of physics, but not all fictional worlds with all imaginable laws of physics.

There are philosophers, such as David Lewis, who have given their interpretation of what the set of possible worlds ought to be. Lewis thinks it should be the set of all spatiotemporal arrangements of "objects" (although this is still a bit vague and makes some assumptions about the structure of spatetime). But Lewis is not the only one with ideas on what the set of possible worlds should be, so the question remains, why should it be that set, and not some other one?

Anyway, to summarize, it's not more parsimonious to be a modal realist about the set of possible worlds usually discussed by modal logicians, because that set of worlds is "too big"; it doesn't help reduce the MDL because it doesn't constrain the possible outputs enough (or at all). But it could be more parsimonious to be a modal realist about a restricted multiverse consistent with certain equations of physics, as in the MWI of quantum physics.