1 Introduction

Lewis’ (1984) Modal realism is the view that for any way the world could be, there is a world that is that way; that each world exists as much, and in the same way, as the actual world; and each world is causally independent of each other world.

Some philosophers worry that if modal realism is true, you have no reason to prevent evils. For if you prevent an evil, you’ll have a counterpart somewhere that allows a similar evil. And if you refrain, your counterpart will end up preventing the relevant evil. Either way one evil is prevented and one is allowed. So if modal realism is true, your act does not make a difference. But it does. So, modal realism is false. Call this ‘the Indifference Argument’. The premise that If modal realism is true, then your act does not make a difference is accepted by proponents and opponents of modal realism alike. As Adams (1974) puts it:

I think that our very strong disapproval of the deliberate actualizing of evils similarly reflects a belief in the absolutely, and not just relatively, special status of the actual as such. Indeed, if we ask, “What is wrong with actualizing evils, since they will occur in some other possible world anyway if they don’t occur in this one?”, I doubt that the indexical theory can provide an answer which will be completely satisfying ethically.

As Lewis (1984) puts it:

If modal realism makes a problem for anyone, it is for utilitarians…. The problem belongs only to utilitarians of an especially pure sort. Only if morality consists of maximizing the total good, absolutely regardless of where and to whom the good may accrue, can it lose its point because the sum total of good throughout the plurality of worlds is contingently fixed and depends not at all on what we do.

As Heller (2003) puts it:

No matter what I do, every ill that a person could suffer, someone will suffer, every evil that a person could do, someone will do, and every good that a person could do, someone will do. My actions can make no difference to the overall pattern of good and bad that happens to people. So there seems no reason for me to bother to do good rather than bad.

As Fischer (2017) puts it:

After all, however I act, I have a counterpart that does the opposite. So if in my moral deliberations I attend to those who aren’t my world-mates, then my counterpart ignores them. And as Heller points out, there’s no good reason for me to prefer that I act rather than my counterpart.

I think these authors are mistaken to be so confident that your act does not make a difference. If modal realism is true, you are in a variant of Newcomb’s Problem. And what you should say about whether the objection is successful depends on what you should say about Newcomb’s Problem. If one-boxing is true, then the problem is a serious one. Or, at the very least, the premise in question is true. And the relevant authors are right to assume it. But if two-boxing is true, then the sort of argument given for two-boxing may be appropriated in favor of saving the child, given modal realism. So I think there is a parallel between the two problems. And it is the burden of this paper to draw the relevant parallel out.

Why care about this? What is the interest of establishing a parallel between the two problems? The parallel is relevant from the perspective of systematic philosophy. Lewis was a two-boxer. Suppose I am right about the parallel. Then, as an historical issue, it is interesting to see that there is a solution to this problem for Lewis’ view, that it comes naturally out of Lewis’ other views, and that it is a solution Lewis overlooked.

The parallel is also relevant because two-boxing is the opinion of the majority in the literature on Newcomb’s Problem. One-boxing has its able defenders. But two-boxing is the dominant view. Everyone in the literature on the indifference argument takes the relevant premise for granted. But, if I am right that there is a parallel between Newcomb’s Problem and modal realism, then the modal realist has new and underexplored resources to address the indifference argument.

I myself am agnostic between one-boxing and two-boxing. So for that reason I am unsure how far this line can be pushed on behalf of modal realism. But for someone like Lewis who is confident of two-boxing, I think it is plausible that the sort of line taken to defend two-boxing can also be used to defend modal realism from the indifference problem.

2 The Structure of the Paper

In Sect. 3, I describe Newcomb’s Problem. In Sect. 4, I offer my initial defense of the claim that there is a parallel between Newcomb’s Problem and your predicament given modal realism. In Sect. 5, I argue that one parallel is that saving the child dominates allowing to die just as taking two boxes dominates taking one box. In Sects. 6, 7 and 8, I consider complications involving the application of Lewis’ analysis of counterfactuals to counterpossibles. In Sect. 6, I discuss the problem. In Sect. 7, I argue that the natural way of extending Lewis’ theory to counterpossibles yields support for the parallel. In particular, whether your counterpart saves, given modal realism, is causally and counterfactually independent of whether you save. This is just as whether there is one thousand in the opaque box is causally and counterfactually independent of whether take both boxes. In Sect. 8, I address concerns ones might have about how the closeness relation works in the context of impossibility. In Sect. 9, I raise and address a series of objections.

3 Newcomb’s Problem

Imagine you are presented with two boxes. One contains one thousand dollars. The other contains either one million or nothing. You may take either both boxes or only the second. Before your decision, a perfectly reliable prediction is made. If the predictor determines that you will take both boxes, she will place nothing in the second box. If the predictor determines that you take only the second box, it will contain one million. Following Lewis (1979, pp. 237–238), note that it does not matter how she makes her prediction. All that matters is that there is a process, such as simulation, that is causally independent of your decision and that could be used to predict what you will decide. One salient feature of Newcomb’s Problem is this: whether you have as an alternative the best outcome depends on what your causally independent replica does. In particular, whether you have a better alternative—a thousand plus a million—rather than a worse alternative-a thousand without the extra million-depends on what your replica does.

In Newcomb’s Problem, evidential and causal difference-making come apart. One way to understand difference-making is as providing evidence (Ahmed and Price 2012; Horgan 1981; Horwich 1987). And in that case you should pick one box. Your replica makes decisions in the same way that you do. So if you pick one box it will provide you with evidence that she picked one box too. And if you pick two boxes, you will have evidence that she picked two boxes.

Another way to understand difference-making is as causing (Lewis 1979, 1981; Gibbard and Harper 1981). In that case, you should pick two boxes. For your replica’s decision is causally independent of yours. Either she picked two boxes or she picked one. You have no control over that. So whatever she did, you will be one thousand richer if you take both boxes and one thousand poorer if you take one.

The solution to Newcomb’s Problem Lewis endorses, and favored by most philosophers,Footnote 1 maintains that causal-difference making, and not evidential difference-making is what matters.

4 Modal Realism is a Newcomb Problem

Now go back to your predicament given modal realism. You have to decide whether to intervene to prevent evil or not. And you know that however you act your counterpart will do the opposite. Your predicament, given modal realism, is exactly like your predicament in Newcomb’s Problem with one irrelevant exception: In Newcomb’s problem your replica acts however you act. Given modal realism, your counterpart acts however you don’t act. In both cases, your behavior can be predicted by your replica’s behavior. And in both cases, your replica’s behavior is causally independent of yours. So if you intervene to prevent an evil, you will have evidence that your counterpart refrains. And if you refrain, you will have evidence that your counterpart prevents. But what you do does not causally influence what your replica does.

If we accept the standard solution to Newcomb’s Problem, then we can answer the Indifference Argument. Providing evidence for an outcome is irrelevant to decision-making. What matters is causally influencing an outcome. Given modal realism, your counterpart in the other world is causally independent of you. She will either prevent or allow. You have no control over that. The one thing you can control is whether you prevent or allow. And you are obligated to prevent.

5 Domination and the Parallel

We can bring out the parallel I have in mind by thinking in terms of domination. Start with Newcomb’s Problem and why two-boxing dominates one-boxing. Consider:

 

Predictor places $1,000,000 in opaque box

Predictor places nothing in opaque box

Two-Box

$1,001,000

$1000

One-box

$1,000,000

$0

There are two situations you might be in. In one, the predictor places one million in the opaque box. In the other, the predictor places nothing in the opaque box. In either situation, you are better off taking both boxes. So two-boxing dominates one-boxing. Now let’s revisit your predicament given modal realism. Consider:

 

Otherworldly counterpart saves

Otherworldly counterpart allows to die

You save

Two saved

One saved and one dead

You allow to die

One saved and one dead

Two dead

There are two situations you might be in, given modal realism. In one, your counterpart saves. In the other, your counterpart allows to die. In either situation, you are better off saving the child. So saving the child dominates allowing to die in modal realism just as picking two boxes dominates one boxing in Newcomb’s Problem.

So, I claim, one relevant feature of the two cases that suggests the Parallel is this: In Newcomb’s problem picking two boxes dominates picking one box. In modal realism, saving the child dominates allowing the child to die.

6 Lewis’ Analysis of Counterfactuals and the Parallel

A relevant feature of Newcomb’s Problem is that whether the thousand is in the opaque box is determined by a process that is causally independent of what you choose. So you might as well take both boxes. Whatever is in the other box you will get an extra hundred without losing a thousand if you do. This works because the following counterfactual is true:

C1: If you were to take both boxes, then the opaque box will not be empty when it would otherwise have contained one thousand.

I claim there is a parallel between the two problems because whether the child in the other world is saved by your counterpart is determined by a process that is causally independent of what you do. So you might as well save the child. Whatever your counterpart does at the other world, the plurality will get an extra saved child if you do. So the parallel works only if the following is true:

C2: If you were to save the child, then no child in the plurality or worlds will be allowed to die that would otherwise have been saved.

For Lewis (1968), a counterfactual ‘if A were true, then C would be true’ is true at a possible world W iff there is a world at which A and C are true that is closer to W than any world at which A is true and C is false. There is a question about which sort of similarity relation is relevant in the analysis. Lewis (1979) holds that in standard contexts the relevant similarity relation is governed by the following weighted priorities:

  1. (1)

    It is of first importance to avoid big, widespread diverse violations of physical law.

  2. (2)

    It is of second importance to maximize the spatiotemporal region of perfect match of particular fact.

  3. (3)

    It is of the third importance to avoid small, localized simple violations of physical law.

  4. (4)

    It is of little or no importance to secure approximate match to particular fact.

Given this analysis, it is clear why C1 is true. In order for the contents of the opaque box to be empty when it would have contained one thousand, there would have to be at least a small, localized simple violation of physical law. If the box is already empty, or if the box has one thousand and continues to have one thousand, then no violation of law occurs. So the closest worlds at which you take both boxes are ones in which you do not miss out on one thousand that you would have otherwise gotten.

But now consider C2. The value for A, in this case, is about you in one world. And the value for C, in this case, is about the entire plurality of worlds. There is no world that contains the plurality of worlds. And there is therefore no world in which you save the child at one world and your counterpart’s act in another world is unchanged that is closer to the actual world than any other world. Thus, C2 is necessarily false.

7 The Krakauer Analysis of Counterpossibles and the Parallel

Modal realism has the resources to admit impossible worlds. And with impossible worlds added to the Lewisian ontology and with Lewis’ (1979) similarity relation extended to contexts involving impossibility in the right way, C2 comes out as true.

Following Krakauer (2012, 2013), impossible worlds are built out of sets of structured propositions. A structured proposition is an ordered set composed of possible worlds, their parts, and relations. A set of structured propositions can be used to form structured worlds. A structured proposition is true at a structured world iff it is a member of the set that composes that world. An impossible world is a structured world that is not complete and consistent. Unlike possible worlds, impossible worlds are so varied and may become so strange that they share no further interesting logical structure.

Impossible worlds describe ways the world could not be. One such way is this: It could not be that there is a world containing the plurality of worlds. And it could not be that there is a world with you at one world and your counterpart at a different world. Thus, there are impossible worlds at which you save (refrain) and your counterpart’s act is saves (refrains).

With impossible worlds added to Lewis’ ontology, and with some such worlds containing both you and your counterpart and the plurality of worlds, the next question is what similarity relation governs contexts involving impossibility. Krakauer (2013) extends Lewis’ (1978) account of closeness to include the following weighted priorities:

  1. (1)

    It is of first importance to minimize the degree of violation of logical laws.

  2. (2)

    It is of second importance to minimize the degree of violation of metaphysical laws.

  3. (3)

    It is of third importance to avoid big, widespread diverse violations of physical law.

  4. (4)

    It is of fourth importance to maximize the spatiotemporal region throughout which perfect match of particular fact prevails.

  5. (5)

    It is of fifth importance to avoid even small, localized, simple violations of physical law.

  6. (6)

    It is of sixth importance to avoid violations of the theorems of the best system that are not laws.

  7. (7)

    It is of little or no importance to secure approximate similarity of particular fact.

Now consider again the counterfactual that interests us:

C2: If you were to save the child, then no child in the plurality or worlds will be allowed to die that would otherwise have been saved.

As we have already observed, there are the impossible worlds at which you save and your counterpart allows to die when she otherwise would have saved. And there are the impossible worlds at which you save and your counterpart saves or allows to die just as she would have if you had allowed to die. C2 is true iff one of the impossible worlds where you save and your counterpart acts just as she otherwise would have that is closer to the actual world than any of the worlds at which you save and your counterpart allows to die when she otherwise would have saved.

Now, what are the ways in which your counterpart might allow to die when they otherwise would have saved? One way is that there is an occult non-causal necessary connection between how you act and how your counterpart acts. Worlds at which your counterpart allows to die in this way score badly with respect to violating metaphysical law. For Humean recombination holds that there are no necessary connections between distinct existences. Worlds at which your counterpart just does whatever they would have done if you had allowed to die do not violate metaphysical law in this way. So they score better with respect to (2).

Another way in which your counterpart might allow to die when she would otherwise have saved is if there is a causal connection between your act and her act. But worlds at which your counterpart allows to die in this way score badly with respect to violating logical law. For the scenario in which you causally impact someone in a causally isolated world is contradictory. Worlds at which your counterpart just does whatever she would have if you had allowed the child to die do not violate logical law in this way. So they score better with respect to (1). On the other hand, one could suppose this occurs at an impossible world at which worlds are not causally isolated from other worlds. But that would be a violation of metaphysical law. And so would incur a violation of (2).

In all other respects, the two sorts of worlds are equal. To get your counterpart to allow to die when she would otherwise have saved would require violating logical or metaphysical law. Thus, there is a world at which you save and no child in the plurality is allowed to die that would otherwise have been saved that is closer to the actual world than any world at which you save and a child in the plurality is allowed to die that would otherwise have been saved. So C2 is true. And therefore Lewis’ analysis of counterfactuals, when extended in the natural way, vindicates the Parallel.

So, presuming the truth of two-boxing, another relevant feature of the two cases that supports the Parallel concerns difference making: In Newcomb’s Problem, the counterfactuals come out in such a way that your act does not make a difference to what the predictor does. But it does make a difference to whether you get an extra hundred. In modal realism, I claim, the counterpossibles come out in such a way that your act does not make a difference to whether your otherworldly counterpart saves. But it does make a difference to whether an extra child in the plurality is saved.

8 Concerns About Closeness Between Impossible Worlds and the Parallel

One complication for the parallel I am raising concerns the otherworldliness of your counterpart given modal realism. This requires us to go to bizarre impossible worlds and to adopt an extension of Lewis’ analysis of counterfactuals. So I should say a bit to assuage worries that the complications arising from this extension will pose a problem for the Parallel.

8.1 The Objects of Comparison Are Too Different

One might wonder how to measure the closeness between a causally disconnected impossible world whose causally connected components represent possible worlds and the actual world. The objects being compared are drastically different. How does one measure the similarity between a single peanut and a jar of peanuts?

Imagine I am holding a peanut. Imagine in front of me there is a table. On the left sits a single peanut. On the right sets a collection of peanuts. Imagine I want to know whether the peanut on the left or the collection on the right is most similar to my peanut. I determine this by specifying the respects of similarity that I care about and checking to see whether the peanut on the left or the collection on the right best matches my peanut in those respects. Imagine, for example, that the peanut in my hand is salty and roasted. And that saltiness and roastedness are the respects of similarity that I care about. If it turns out that the single peanut on the left is unsalted and unroasted, and if it turns out that the collection on the right is salted and roasted, then the collection on the right is more similar to my peanut in the relevant respects than the peanut on the left. In this way, one can measure similarity between a single peanut on the one hand, and a collection of peanuts, parts of peanuts, and relations between them, on the other.

In the same way, one can measure whether a single world or an impossible world that contains a plurality of worlds is closer the actual world. Pick the respects of similarity that you are interested in. Suppose we are interested in whether the single world or an impossible world containing a plurality of worlds is more similar to the actual world with respect to avoiding violation of logical law. Now consider two impossible worlds:

W1: A single impossible world exactly like the actual world with the exception that I am wearing a contradictory t-shirt.

W2: An impossible world that is a collection of a plurality of worlds. None of the worlds in the collection are contradictory. Nothing about the collection is contradictory.

By the relevant measure of similarity, W2 is closer to the actual world than W1. Even if a collection of worlds, their parts, and relations between them is a different sort of thing than a single world, they can nevertheless be compared for their similarity to the actual world just as a single peanut can be compared for its similarity to my peanut with a collection of peanuts.

8.2 Differences Between the Logic of Possibility and Impossibility

One might worry that Krakauer’s account does not match the logic of Lewis’. An account of how Krakauer’s theory of impossible world works with Lewis’s modal realism is needed. In particular, consider the subjunctive conditional

C′: If you were to save the child, then you would feel good about yourself.

C′ and C2 share the same antecedent. C′ is to be evaluated at the nearest possible world where you (or your counterpart) save the child, but according to Krakauer’s extension of Lewis’ theory, C2 is to be evaluated at a causally disconnected impossible world that’s nearest to the actual world according to some criteria of nearness. Given that they have the same antecedent, how is it that C′ is to be evaluated by comparing only possible worlds but C2 is to be evaluated by comparing impossible worlds to the actual world?

Recall that Lewis’ theory is this:

A counterfactual ‘if A were true, then C would be true’ is true at a possible world W iff there is a world at which A and C are true that is closer to W than any world at which A is true and C is false.

C′ and C2 share an antecedent. But they have different consequents. In the present context, having different consequents explains why C′ is to be evaluated by comparing possible worlds and C2 is to be evaluated by comparing impossible worlds. Consider C2: As was observed in motivating objection six, there are no possible worlds at which there are truths or falsehoods about the plurality of worlds. So there is no possible world at which you save the baby and then there is any proposition about the plurality of worlds that is true or false. Only impossible worlds have such truths and falsehoods. So that is why, in the case of C2 we need to go to impossible worlds to evaluate the relevant counterfactual. Contrast this with C′: All the possible worlds have truths about whether you feel good or not. So the closest world at which you save and feel good and the closest world at which you refrain and do not feel good are still possible worlds. Thus, although C′ and C2 share antecedents, it is the difference in consequents that require going to impossible worlds when evaluating C2 but not C′. For truths and falsities of propositions such as ‘you feel good’ are had by impossible and possible worlds alike. But truths and falsities of propositions such as ‘one extra child in the plurality of worlds dies’ are had only by impossible worlds.

8.3 Impossible Worlds Cannot Represent the Plurality

Regardless of the criteria for closeness, one might think it is pointless to evaluate C2 at an impossible world that doesn’t represent the plurality. We evaluate C′ at a world where you or your counterpart save the child. According to Krakauer, to evaluate ‘If nine was prime, then it would not be divisible by three’, we evaluate at an impossible world where nine is prime. If we are to evaluate C2, there doesn’t seem to be any reason to look at a scenario where the boundaries of possibility were drawn differently. We should always evaluate C2 and similar statements at the impossible world that represents the plurality.

I don’t think this is quite right. There is just one Hubert Humphrey in all of the plurality of worlds. And that one Humphrey lost the election. But we still think Humphrey could have won. For Lewis, the way we analyze whether Humphrey could have won is not by considering Humphrey himself. Rather, we go to a different world where a counterpart of Humphrey represents him as winning. That counterpart is not a perfect replica of Humphrey. But it nevertheless represents him. And if we want to know what Humphrey could have been like, and to evaluate counterfactuals about him, we don’t look at Humphrey himself, we look at his counterparts in other worlds.

In the same way, just as statements about what could have happened to Humphrey are to be evaluated by considering counterparts of Humphrey that are not identical to him but represent him, statements about what could have happened to the plurality of worlds are to be evaluated by considering counterparts of the plurality that are not identical to it but that represent it. If we want to know what the plurality of worlds could have been like, if we want to evaluate counterpossibles about it, then we do not look at the plurality of worlds as it is. We go to impossible worlds that are not identical to the plurality but that represent the plurality.

In metaphysics, it is standard to consider scenarios in which the boundaries of possibility are drawn differently. The mind body dualist may consider how things would have been if materialism were true. The materialist may consider how things would have been if dualism were true. At most one of these views is right. And the other merely describes impossibilities. To talk about how the mind might have been if materialism were true, we must go to impossible worlds in which counterparts of the mind represent it but are not identical to it. To talk about how the pluriverse might have been if it were different, we must go to impossible worlds in which counterparts of the plurality of worlds represent it but are not identical to it.

9 Objections and Replies

First Objection: It is inevitable that if you save your counterpart will refrain. And if you refrain your counterpart will save. Following Heller (2003, p. 13), it is the inevitability relation and not the causal relation that matters. As Heller puts it “Why think it is the causal connection that is relevant…? [In some cases] the inevitability is grounded in a casual relation, but the grounding does not seem to me to matter. What matters is the inevitability, regardless of why it is inevitable.”

Reply: Notice how similar Heller’s reasoning is to Leslie’s (1991, p. 397) argument for one-boxing:

Crossing a plain, I come to what looks like a gigantic mirror. But pushing a hand against it, I feel flesh and not glass. The universe is symmetric, the flesh that of my double… [and] fully deterministic…. I control what my double does…. Not, of course, of being able to guarantee causally, but rather of being able to ensure, guarantee, in the fairly straightforward… sense that if I were to throw stones, then he would…. It is just as good as true causation for ensuring that things get done. If I want a bird killed in the other universe-half, all I need do is throw a stone appropriately in my half.

Two-boxers, such as Lewis, deny that inevitability is sufficient in evaluating Newcomb’s Problem. They maintain that causal inevitability and not just any inevitability at all is what matters. So what the modal realist should say in response to Heller on the point in question is this: It matters what grounds the inevitability because the standard solution to Newcomb’s Problem implies that it matters. The solution to Newcomb’s Problem that Lewis endorses holds that causal inevitability but not evidential inevitability matters. Given modal realism, you are in a Newcomb’s Problem. So causation is what matters in your case as well. Or at least, that is what the two-boxer should say.

Second Objection: Modal realism posits a necessary logical connection between your act and your counterpart’s act. In Newcomb’s Problem, there is merely a contingent evidential connection between your act and your counterpart’s act. Necessary connections make a difference even though mere evidential connections do not. So whereas your act in Newcomb’s Problem does not make a difference, your act given modal realism does.

Reply: Modal realism does not posit a necessary connection between your act and your replica’s act. Indeed, Lewis explicitly endorses Humean recombination. This view “denies necessary connections between distinct existences.” You and your replica are distinct existences. So there is no necessary connection between the two of you. The only connection is evidential. You know prior to your act that the possible worlds are arranged in such a way that you and your counterpart will act in opposite ways. So if you save, you will have decisive evidence that your counterpart refrains. And if you refrain, you will have decisive evidence that your counterpart refrains. But there is not some causal or logical power you have to make your replica act in one way or another. And so given modal realism, you are in a variant of Newcomb’s Problem and you are obligated to save.

Third Objection: The lesson of Newcomb’s Problem is that counterfactuals matter. Causal difference making matters only insofar as it is accompanied by a counterfactual difference. In Newcomb’s Problem it is. But in Lewis’ theory it is not.

Reply: You do make a counterfactual difference. Your counterpart either does good or does evil. You have no control over that. But whatever your counterpart does, you have control over what happens at your world and the contribution your world makes to the plurality of worlds. However your counterpart acts, you will add one saved life to the plurality if you save and you will add one death to the plurality of worlds if you refrain. Of course, after you do good you will have evidence that your counterpart has done evil. But your counterpart would have done evil no matter what you did. So if you had done evil rather than good, the plurality of worlds would have ended up worse.

Fourth Objection: Look at all the people throughout the plurality of worlds who act in one way. Their counterparts in the other world act in the other way. So there is a counterfactual connection between your act and your counterpart’s. If you do the right thing, you would have done just as well for the pluriverse as a whole to do the wrong thing.

Reply: This is a variant of the Why Ain’cha Rich? Argument. Evidential decision theorists point to all the people who previously encountered Newcomb Problems and point out that all the one boxers are rich and all the two boxers are poor. As Lewis (1981, p. 377-8) puts it:

The one-boxers sometimes taunt us: if you’re so smart, why ain’cha rich? They have their millions and we have our thousands, and they think this goes to show the error of our ways. They think we are not rich because we have irrationally chosen not to have our millions. We reply that we never were given any choice about whether to have a million. When we made our choices, there were no millions to be had. The reason why we are not rich is that the riches were reserved for the irrational.

Similarly, given modal realism, it will turn out that those who act wrongly will have counterparts that act rightly. And those who act rightly will have counterparts who act wrongly. But this is not because the act of your counterpart counterfactually depends on your act. Instead, there is only evidential dependence. If those who acted wrongly had acted rightly instead, that would not have changed whether their counterparts acted rightly. And if those who acted rightly had acted wrongly, their counterparts would have acted wrongly too. Lewis should reply that the plurality of worlds is set up in such a way that those who do good were never given any choice about what their counterparts did. Counterparts who do good are reserved for those who choose evil. The standard view is that such considerations do not favor picking one box in Newcomb’s Problem. If that is right, they do not recommend indifference given modal realism.

Fifth Objection: It is mysterious how the possible worlds could be arranged so that you and your counterpart will act in different ways without there being a necessary connection between you.

Reply: It is mysterious. But it is also mysterious that there is an occult non-causal necessary connection between you and someone else in another possible world. Either way there is a mystery for Lewis. Why not go with the mystery that does not contradict the core doctrine of Lewisian metaphysics—Humean recombination? Why not go with the mystery that does not lead to the indifference objection?

Furthermore, imagine someone objected to Lewis’ modal realism in the following way: It is mysterious how the possible worlds could be set up in just the right way to make our modal claims true. Lewis would reply that the reason to believe modal realism is because of its explanatory power. It explains a bunch of metaphysical things that seem true. But if that is enough to believe the worlds are arranged in such a way as to make them true, then it is enough to believe that worlds are arranged in such a way that makes our moral beliefs true. If it isn’t enough, then the indifference objection is not an independent objection to modal realism.