What Are Possibilities? Part 3B: More Objections to EMR

In last week’s post, we discovered that Lewis’ theory of Extreme Modal Realism (EMR) faces four key objections. First, Lewis reduces actuality to an indexical claim about which entities are spatio-temporally related. Second, in order to reduce the modal to the non-modal, he invokes a counterpart relation which is just as mysterious as the concept of modality he is attempting to explain. Third, his theory has the odd result that causally irrelevant facts are involved in making ability-statements true. And fourth, EMR conflicts with the intuition that the sum total of all real worlds might have been different in certain respects – say, by being fewer in number. For this week’s post, let’s continue with some additional objections…

time_space
Extreme Modal Realism implies that non-spatial entities like God cannot possibly exist

Fifth, EMR denies that concrete non-spatial entities (e.g. spirits, God, etc) are even possible (Pruss, 1996, p.116). Why? Because Lewis defines a possible world as collection of concrete entities that are spatio-temporally related to each other, but to nothing else. Non-spatial entities are thereby excluded.  However, while it is true that many philosophers deny the existence of non-spatial entities, even the most convinced atheists do not wish to burden themselves with the stronger claim that God could not possibly have existed.[1] Perhaps Lewis can avoid this consequence by relaxing his conditions for membership in a world – say, by arguing that entities are “world-mates” only if they are capable of causally interacting with each other. But if the world-mate relation is understood in this way, then Lewis is caught in a vicious circle. His attempt to explain the modal in terms of the non-modal requires that his world-mate relation be defined non-modally as well. And that’s just the problem, because causality is a modal notion! Admittedly, Lewis tries to reduce causal truths to (non-modal) counterfactual truths (Pruss, 2011, pp.42-52), but recall that his account of counterfactuals already presupposes that worlds can be compared in terms of similarity, and this in turn assumes that the question of membership in those worlds has already been settled – which it hasn’t.

Sixth, EMR is too revisionist about ethics, regardless of whether one interprets Lewis’ theory along the lines of counterpart or identity theory. For various reasons, Lewis prefers a deontological framework of ethics which emphasizes the intrinsic features of an action (not its consequences) as the main determinants of its moral status. Pruss (1996) agrees with Lewis preference for deontology, but he also notes that consequences sometimes matter within such a framework: “While with Lewis I reject consequentialism as a general account of morals, it is true that consequentialistic reasons sometimes have a rightful place in our deliberation” (p.126). As it turns out, Lewis’ view runs into problems if it allows consequences to be relevant in this way.

In the movie "Sliding Doors" (1998), Gwyneth Paltrow's character branches off into two alternate futures. Perhaps both Gwyneths are the same person in each future (the identity view), or each Gwyneth is a distinct, yet similar individual (the counterpart view).
In the movie “Sliding Doors” (1998), Gwyneth Paltrow’s character branches off into two alternate futures. Perhaps both Gwyneths are the same individual in each future (the identity view), or each Gwyneth is a distinct, yet similar individual (the counterpart view).

To understand why, consider what happens if we assume the identity view (not Lewis’ preferred option). If the identity theory is true, then the individual instances of me in other worlds are me!  And if they are me, then my duties are not locally constrained to this world because the actions I perform here make a difference to whether I fulfill my duties in other worlds. For example, my refraining from murdering a stranger named Bob has the result that I do murder Bob in another world! No matter what I do, the result is that I become an evildoer, since the other worlds in which I murder Bob are just as real as this world. Thus, in the grand scheme of things it does not really matter whether refrain from murdering Bob or not, because I always fail to fulfill my duty in some world or other. As Pruss (1996) asks, “must one strive hard to improve some virtue if one knows that this will result in one not having that virtue…in another world? That would seem a little questionable” (p.122).

doubt1
David Lewis’ modal ontology reduces some of our basic moral intuitions to matters of indifference

Similar ethical problems follow from a counterpart interpretation of EMR. Why? Because it conflicts with the strong “moral intuition that the lives of all existent strangers in need who are in objectively similar circumstances are of equal significance” (Pruss, 1996, p.130). To clarify this intuition, consider one of Pruss’ many thought experiments:[2] suppose (as most people do) that I have a prima facie moral duty to perform a trivial action to save a stranger’s life – call him, Bob.[3] If the counterpart theory is correct, then this duty is undermined because if I save Bob’s life, then his counterpart in another world dies anyway. I may not cause Bob’s counterpart to die (because Lewis’ worlds are incapable of interacting with each other causally) but it is still the case that were I to save Bob’s life, a counterpart of his would die – or were I to refrain, a counterpart of Bob’s would live.[4] Since both Bob and his counterparts are real persons, the ethical rationale for saving one rather than the other is undermined because either way someone dies. Thus, the counterpart interpretation EMR means that saving Bob is a matter of indifference.[5] But surely, our ordinary moral intuitions tell us that such an action is not morally indifferent.

Lewis’ view leads to skepticism about inductive inferences

Seventh, EMR leads to skepticism about induction. Induction is a way making general inferences from particular instances of events. If a sufficient number of instances have the same pattern, then we can infer that future instances will likely exhibit the same pattern. One way of understanding these instances is to say (with David Hume) that they are constant conjunctions of events. That is, if we regularly observe A’s being conjoined with B’s, then it is probable to suppose that events of type-A and type-B will also be conjoined in the future. Unfortunately, if induction is nothing more than generalizing from regular conjunctions, then our predictive abilities are limited indeed. For given any finite number of A’s followed by B’s, it will always be possible that some future A will not be followed by B. At most, we can know with a high degree of probability that a finite number of A’s and B’s will be conjoined in the future – not that all of them will.[6]

But the situation is even worse for Lewis’ modal realism. Why? Because within the vast array of concrete worlds that exist in Lewis’ ontology, there are just as many worlds in which induction holds as those in which it fails. Assume for the sake of argument that the actual world contains 1000 past instances of A/B conjunctions prior to time t1. It follows from EMR that there are just as many (if not infinitely many) worlds that contain the very same history of 1000 instances prior to t1, but fail to have A/B conjuncts at t2, t3, and so on. But if there are equally many worlds in which induction succeeds or fails after t1, then we can’t know with any probability above 0.5 whether any such world (including the actual world!) will be friendly to induction.

Lewis might reply to this objection by claiming that for inductive inferences to be justified in the actual world, we only need to take its own unique past history into account, not the same histories in other worlds. But this response will be of not help to Lewis, because these histories are just as real (and therefore just as relevant) to what we infer. So it seems after all that EMR leads to skepticism about induction.


[1] This stronger claim is even more striking upon further reflection because it entails that God could not have existed, even in a possible world that contains no suffering!

[2] In each thought experiment, Pruss (1996) asks us to imagine a possible world “where the laws of nature and [initial] conditions allow for one and only one free and indeterministic choice, everything else being deterministic and there being no other choices (deterministic or not) nomically possible” (p.123). He insists on this constraint because he wants to make sure that temporal considerations such as the formation of virtuous or vicious character in the future do not influence the process of moral deliberation in these thought experiments.

[3] Pruss uses the example of a stranger because if Bob was my friend, then due to the moral obligations that issue from the social significance of friendship, I may be obligated to save his life rather than his other-worldly counterpart, because I am not friends with his counterpart.

[4] Interestingly, the idea that counterfactuals hold between worlds may be threatening to Lewis’ theory insofar as he reduces causation to counterfactuals. At the very least, Lewis would need to argue that such transworld counterfactuals are non-causal, or that, despite appearances, there are no transworld counterfactuals.

[5] A similar kind of moral indifference undermines the basis for encouraging beneficial but non-obligatory actions, since again, there are infinitely many worlds in which the benefit obtains or does not obtain, resulting in an overall net benefit of zero.

[6] For a defence of the claim that constant conjunction can only justify inferences about a finite number of future instances, see Hume’s Abject Failure: the Argument Against Miracles, by John Earman (2000), p.30.

One thought on “What Are Possibilities? Part 3B: More Objections to EMR”

  1. It should be noted as well that Lewis is committed to the existence of *sets*, since he defines properties, propositions, and possible worlds in terms of sets of concretely existing things. But if sets are abstract (non-spatial) objects, then according to the fifth objection above, it is false that only spatial entities can be world-mates.

Leave a Reply

Your email address will not be published. Required fields are marked *