Introduction
G. A. Cohen has argued that there is a surprising truth in conservatism—namely, that there is a reason for some valuable things to be preserved, even if they could be replaced with other, more valuable things. This conservative thesis is motivated, Cohen suggests, by our judgments about a range of hypothetical cases. After reconstructing Cohen's conservative thesis (section 1), I argue that the relevant judgments about these cases do not favor the conservative thesis over standard, nonconservative axiological views (section 2). But I then argue that there is a case, Mirrored Histories, that is such that, if one shares Cohen's conservative attitude, judgments about this case favor Cohen's conservative thesis over a wide range of non-conservative axiological views (sections 3–4). Reflection on this case also suggests a different explanation of apparently conservative judgments that merits consideration in its own right. Taken together, my arguments help to clarify the commitments of Cohen's conservative thesis and the sources from which it might draw undefeated intuitive support.
1. The Conservative Thesis
Cohen writes that there is a reason ‘to conserve what is valuable, that is, the particular things that are valuable’ (2012: 152). My focus is the view that Cohen describes in the quoted sentence.Footnote 1 In order to draw out its distinctiveness, I need to give a definition and draw a distinction. I will say that a thing o is conserved at a time t iff at t, o exists and, if it existed prior to t, is not significantly altered from its condition immediately prior to t (Cohen Reference Cohen and Otsuka2012: 147). So, on this definition, one way in which something could fail to be conserved at a time is by being destroyed at that time. In most of what follows, that will be the only salient way in which some thing fails to be conserved. For remarks on the case of significant alternation, see the end of this section.
I also need to distinguish between two different facts. Suppose there is some finally valuable thing o—that is, some thing o that is valuable for its own sake. (For short, in what follows, I sometimes abbreviate ‘finally valuable’ to ‘valuable’.) On the one hand, there is the fact that the world contains o—that is, the fact that o, a particular bearer of value, exists. On the other hand, there is the fact that the world contains the value of which o is the bearer—that is, the fact that a certain amount or kind of value, borne by o, exists.
With this definition and this distinction in hand, I can now state a preliminary version of Cohen's conservative view as I shall understand it:
(C1) For any finally valuable thing o: if o exists, then the fact that o exists is a pro tanto reason for o to be conserved.
The distinctive, surprising claim expressed by C1 is that the mere fact that o is an existing bearer of value is a reason, over and above the reason given by the value it bears, for o to be conserved. That is, according to C1, there is a reason for the particular things that are bearers of value to be conserved. Thus C1 is not the uncontroversial claim that the amount or kind of value borne by o is a reason for o to be conserved. That claim would be consistent with the axiology of utilitarianism or any version of (nonconservative) pluralist consequentialism. To the contrary, C1 entails the falsity of utilitarianism or any other standard (nonconservative) pluralist consequentialism.
We can next refine C1, the preliminary statement of the conservative view, so that it makes a nontrivial difference to our reasons for action. To illustrate the need for this revision, it will help to introduce a simple representational framework. Let a history of the world be represented by an infinite two-dimensional vector, one dimension corresponding to time and the other to the set of possible valuable objects. Figure 1 illustrates this framework.
Figure 1. Framework
The row that corresponds to each possible valuable object has one of two kinds of entries: a non-numerical symbol (a dash), representing the fact that the object does not exist at a time, or a number that indexes the value of an object at a time period during which it exists. A row that consists entirely of dashes thus indicates the nonexistence of that valuable object in that history; a row that takes a numeric value at any time indicates that that valuable object exists during that time in that history. I will take this framework—adapted from John Broome (Reference Broome2004: 23–26)—for granted, along with the simplifying assumptions that it encodes.
Now suppose someone is to choose, during t1, between the continued existence of one valuable thing o, or the destruction of o and the coming into existence of another valuable thing p. The choice is represented in figure 2.
Figure 2. Conservation or replacement
The desired conservative result is that this person has a reason to favor h1 over h2. But, if we apply C1 as it stands to h2, then, since p exists in h2, C1 entails that there is a reason, given by this fact, favoring h2. To solve this problem, we can remember that, in general, reasons are indexed to times. For example, if I promise to you on Monday to return your book to the library, then I have a reason on and after—but not before—Monday to return your book to the library (cp Vranas Reference Vranas2007: 171). We can claim that, similarly, the reason given by o's existence to conserve o is a reason only at times at which o exists:
(C2) For any finally valuable thing o: if o exists at t, then the fact that o exists is a pro tanto reason at t for o to be conserved.
Reconsider the case represented in figure 2. C2 now implies—the desired result—that this person at t1 has a conservative reason favoring h1 and no conservative reason favoring h2.
Given the argument to come, it will be helpful to note here a different way in which to develop the conservative view. The time-indexing maneuver that generated C2 has the effect, from the point of view of a deliberating agent at some time, of restricting the scope of the view to what then exists. So we can equivalently state C2 as the following view:
(C2) For any presently existing finally valuable thing o: the fact that o exists is a pro tanto reason for o to be conserved.
But some valuable things, although they do not presently exist, will exist no matter what any person does—call those the independently existing things (compare Temkin Reference Temkin2012: 417). We could also have recovered the verdict about the case represented in figure 2 by appealing to this distinction between what is independently existing and what is not. We could have differently restricted C1, revising it instead to the following view:
(C3) For any independently existing finally valuable thing o: the fact that o exists is a pro tanto reason for o to be conserved.
C3, it should be noted, is still tacitly time-indexed. For any object, there is a time at which it becomes settled that it is an independently existing thing—the time after which no agent can affect whether it will exist or not. So C3 can equivalently be stated as:
(C3) For any finally valuable thing o: if o is independently existing at t, then the fact that o exists is a pro tanto reason at t for o to be conserved.
Return now to the case represented in figure 2. In this case, p is not an independently existing thing. So, like C2, C3 entails that the agent in this case would at t1 have a conservative reason that favors h1 but not h2. So this alternative way of restricting C1 would also deliver the desired result.
Because all presently existing things are independently existing, but not conversely, C2 and C3 would sometimes have conflicting implications. Consider a case in which an agent faces a choice during t1 between an independently, but not presently, existing thing being destroyed at a future time or its persisting (Cohen Reference Cohen and Otsuka2012: 216). This choice is represented in figure 3.
Figure 3. Future conservation or destruction
If an agent had to make some small sacrifice of well-being to bring about h4, C3 could imply that she had sufficient reason to do so; C2 could not. C3 would also imply, for example, that there is a reason, if one can, to bring not-yet-existing independently existing valuable things into existence as soon as possible. If we could bring all the people who will certainly exist into existence now, C3 implies we have a reason to do so. And, if it were settled that God was going to create persons, C3 implies that there would be a reason to make them immortal.
I shall not attempt to adjudicate between these two versions of the view. Each could plausibly be claimed to best reflect the judgments that motivate (Cohen-)conservatism. The friend of C2 could claim that the fundamental conservative motivation is a distinctive kind of care about the value that we have. But if a valuable thing does not exist, then it is, in one trivial sense, not part of the value we have. This argument for C2 would not, I believe, be decisive. Just because we happen not to have some value yet, if it is true that we certainly will have it, then it is value that we, in a relevant sense, have (cp. Kagan Reference Kagan2012: 227). The friend of C3 can, moreover, add that, to the extent that intuitive judgments about the case represented in figure 3 are weaker than those concerning the case represented in figure 2, it is because we are disposed to assimilate the comparatively rare class of things that we know are independently, but not presently, existing to the class of things that are dependently, but not presently, existing. As this discussion suggests, C2 and C3 express different views about why conserving valuable things matters. On the former view, the explanation involves our standing in a certain relation (one that minimally involves contemporaneous co-existence) with those things. On the latter view, the explanation does not importantly involve any relation that we shall bear to those things. Without attempting to settle this important question, I work with C2 in what follows. That choice raises a significant extra objection (discussed in section 4.1) to my thesis. If my thesis can be defended even given C2, then so much the better for it if C3 is the better-motivated version of Cohen's view.
Before discussing the case for C2, it is worth making four further clarifying remarks about it. First, I have formulated C2 as a claim about normative reasons. Cohen's conservative view may include claims that go beyond C2. For example, a complete statement of Cohen-conservatism may include claims about the source or ground of the reason referred to in C2. I can afford to frame C2 in this way—remaining neutral about further explanations of the reason referred to in C2—because I am interested only in what Cohen-conservatism implies about our practical reasons, whether these implications can be given intuitive support, and what it would take to do so.
Second, I have formulated C2 as a claim about normative reasons for action. In some contexts, agents cannot make a difference to which outcome eventuates. The case that I describe below in section 3 is one such context. In such contexts, I take it that what would otherwise be a reason to bring about an outcome is, other things equal, a reason to prefer that the outcome eventuates (see Parfit Reference Parfit2011: 47). So, when there is nothing that we can do, I will take C2 to imply that we still can have reason to prefer that certain outcomes eventuate rather than others.
Third, my central argument focuses on cases in which valuable things fail to be conserved at a time because they are destroyed at that time. So I will say little about how my case might be extended to cover an object's failing to be conserved by undergoing significant alteration. But I can briefly illustrate how the framework outlined above could be extended to this case. We could enrich the framework to associate, with each valuable object, several subvectors—that is, several rows—corresponding to the relevant properties of that object, as figure 4 illustrates.
Figure 4. Extending the framework
Significant alteration, in the relevant sense, can then be modeled as the loss or replacement of sufficiently many such properties at a time. I assume that the argument to be given below in section 3 could, mutatis mutandis, apply to a thing's not being conserved through its being significantly altered.
Fourth, Cohen suggests that his thesis applies to those things of which it is ‘not right . . . to ask only what is the best that can be got out of it, or the best that can be made of it’ (2012: 148). If this is correct, the thesis may apply only to a subset of finally valuable things; it may not apply, for example, to pleasurable experiences. I leave this question aside in what follows and make the simplifying assumption that all finally valuable things satisfy the condition quoted above.
2. How Well Do Cohen's Cases Support C2?
In motivating C2, Cohen presents and discusses a range of hypothetical cases. Here is a representative case (adapted from Cohen Reference Cohen and Otsuka2012: 146–47):
All Souls Enhancement. Suppose All Souls College (or the activity that partly constitutes it) is finally valuable. Moreover, suppose that All Souls could be radically changed—e.g., its buildings demolished and replaced by more modern structures, its research and teaching activities diversified, and its fellows replaced—in a way that stipulatively increases its overall final value.
If sufficiently extensive, such enhancements would change All Souls’ ‘central organizing self-conception’ (Cohen Reference Cohen and Otsuka2012: 147). They would significantly alter the kind of thing All Souls is, thereby failing to conserve it. So C2 implies that there would be a reason not to make these changes. If any standard version of (nonconservative) pluralist consequentialism is true, however, there is apparently no reason not to make these changes. So, if one shares Cohen's judgment that there is a reason not to make these changes, then this case seems to provide intuitive support for C2.
However, even if one shares Cohen's judgment about the All Souls Enhancement case, it is questionable whether it provides undefeated intuitive support for C2. Here is the problem. If a proposed set of changes to All Souls is significant enough to alter the kind of thing it is, then these changes would arguably also induce a loss of one kind of final value—namely, the kind of value that is peculiarly instantiated by All Souls’ present good-making features. But, if that is so, then accounting for the reason not to enhance All Souls does not require appeal to C2. It would be enough to appeal to a version of (nonconservative) pluralist consequentialism with a suitably specified axiology that recognizes this kind of value as a basic source of final value. The relevant version of pluralist consequentialism would then imply that there is a pro tanto reason not to enhance All Souls. But this reason would be, not the mere fact that All Souls is an existing bearer of value, but rather the loss of the peculiar kind of intrinsic value of which All Souls is the bearer (Hurka Reference Hurka1996). So Cohen's judgment about the All Souls Enhancement case does not favor C2 over all versions of (nonconservative) pluralist consequentialism.
Similar remarks apply to the other cases that Cohen discusses. Each of these cases concerns the significant alteration or destruction of a similarly complex good—such as great works of art or architecture or other kinds of human achievement. To find undefeated intuitive support for C2, I suggest, conservatives will have to look to a different kind of case. In section 3, I show how they can do so.
One might reply that the version of pluralist consequentialism that would be needed to accommodate Cohen's judgment about the All Souls Enhancement case is implausible. In order to deliver the result that a distinctive kind of value is borne only by All Souls (and not its enhanced replacement), the view in question would seem to have to individuate kinds of value quite finely. But one might think it implausible that kinds of final value are so finely individuated. This might be thought implausible on two grounds. General considerations of simplicity or parsimony disfavor positing many distinct sources of basic final value. Or, at least, one might think that it is, in this context, objectionably ad hoc to posit, successively, a new source of basic value to accommodate each of the various case judgments that C2 putatively explains. It may seem that it is C2, here, that gives the simpler or more unified explanation of this set of judgments. And so C2 may seem to vindicate Cohen's judgment about All Souls Enhancement at least as well as any version of (non-conservative) pluralist consequentialism.
I have four replies to this objection. First, of the reasons to doubt substantive pluralism about value, most are reasons to doubt pluralism as such. Consider, for example, the ‘Why these?’ challenge to pluralist views: Why it should be thought that only some particular number n of candidate sources of final value should happen to be finally valuable (Lin Reference Lin and Fletcher2016: 335)? That challenge arises with equal force for any such n. One challenge, however, might distinctively arise for a pluralist consequentialist view that posits a large number of sources of final value—namely, explaining what the basic sources of value kinds have in common, in virtue of which they would all be kinds of value. This challenge, however, could be defused by adopting a more radical pluralism about value properties themselves. Rather than claiming that there are multiple sources of (the same property of) goodness, one would claim that some things are good-in-the-welfare-way, others good-in-the-equality-way—and one thing good-in-the-All-Souls-way (following a suggestion made by Heathwood Reference Heathwood, Hirose and Olson2015: 142–43 in the prudential-value context).
Second, it is questionable whether C2 could claim to be a more parsimonious or more unified explanation of the data. Admittedly, C2 gives a single explanation of the set of judgments at issue. But it does so only by allowing a disunified, hybrid account of the sources of reasons; some reasons have their source in the value that valuable objects bear, and other reasons have their source in those value-bearing objects themselves. The version of pluralist consequentialism under consideration, by contrast, retains a unified account of the source of reasons; on this view, all reasons have their source in the value that valuable objects bear. Each view is thus simpler in some respects than the other. It is questionable which of these kinds of parsimony matters more, if either matters at all. More generally, when parsimony is not linked to some independent, epistemically relevant consideration—as, for example, in a context in which the simpler scientific hypothesis has a greater prior probability or likelihood—the epistemic relevance of parsimony is in any case questionable (Sober Reference Sober2015: ch. 5). So C2 could not be claimed, on grounds of parsimony, to be an obviously better explanation of the data.
Third, the claim that there is a multiplicity of sources of basic value need not be unmotivated or ad hoc. It is a familiar thought from the literature on prudential value that there is widespread rational regret about, for example, forgoing a lesser pleasurable experience for the sake of another. Substantive pluralism about welfare accommodates this data point by reversing the monistic order of explanation, positing finely individuated kinds of prudential value (such as gustatory pleasure, olfactory pleasure, or visual pleasure) such that a token pleasurable experience is welfare-enhancing because it is a gustatory pleasure, and not—as monists would claim—the reverse. But if there is reason to believe that the sources of prudential value are finely individuated, then there seems little reason to expect the sources of basic value simpliciter to be less finely individuated.
Fourth, in order to accommodate Cohen's judgment about the All Souls Enhancement case, pluralist consequentialism need not posit an extensive list of basic sources of final value. Even a version of pluralist consequentialism that countenances only a small number of basic sources of final value will typically have reason to allow for complex interaction effects among the basic kinds of value that it countenances (Kagan Reference Kagan1988). And this would allow a version of pluralist consequentialism that recognizes only a short list of basic sources of final value to ‘emulate’ a version that posits a lengthier list. (A view that recognizes only five kinds of basic value, each one coming in only three significant quantities, would allow for 243—that is, 35—distinct possible interaction effects.) Thus one might claim that All Souls instantiates a peculiar interaction effect among basic sources of value, and not a distinct kind of basic value. Since Cohen's best examples all involve cases involving complex goods—All Souls College, David Hockney paintings, Michelangelo's Pietà—the appeal to such interaction effects would not, I believe, be unmotivated or ad hoc.
To summarize the discussion to this point: I have not suggested that C2 does not accommodate and explain Cohen's judgment about the All Souls Enhancement case. I have suggested, instead, that there is an alternative explanation available for the judgment, one that appeals only to a version of pluralist consequentialism. And I have given a prima facie case that the pluralist consequentialist view in question would not be obviously implausible, ad hoc, or unmotivated. It does not follow that a pluralist consequentialist view would better accommodate and explain the judgment in question. My suggestion is only that, in light of this possibility, C2 is not clearly the best explanation of this judgment. To the extent that cases like All Souls Enhancement offer intuitive support for C2, then, it is questionable whether the intuitive support they offer is undefeated. That motivates asking whether, by appealing to a different kind of case, conservatives might find undefeated intuitive support for their view. In the next section, I explain how they might do so.
3. The Mirrored Histories Case
Conservatives can find undefeated intuitive support for their view, I suggest, by appealing to the following Mirrored Histories case. Let time extend infinitely, with neither beginning nor end, and let h1 and h2 be two possible histories—that is, two ways in which things might go. In each of these histories, suppose that the very same finally valuable thing, o, will certainly exist, and suppose that no other valuable things ever exist. These two possible histories differ in only one respect. In h1, until some time t, the universe is empty of valuable things; at t, o comes into existence and thereafter never ceases to exist. In h2, the very same (that is, numerically identical) thing o has always existed, until t, at which time it ceases to exist and the universe is thenceforth empty of valuable things. The only difference between h1 and h2, in other words, is when o exists. In h2, the beginning of t is the last moment of o's existence; in h1, its first. Figure 5 illustrates the case:
Figure 5. Mirrored Histories
Suppose also that there is a single person who has just woken up, at the beginning of t, with memory loss. She knows all the above facts. But she does not know whether she is in h1 or h2, and she cannot otherwise affect what happens. According to any standard version of nonconservative pluralist consequentialism (for example, views that recognize overall welfare, inequality, priority to the worse off, autonomy, or perfectionism as basic sources of value), she has no reason to prefer that she is in h1 rather than h2 or vice versa. But, at t, o is presently existing. So C2 entails that, at t, this person has a reason to prefer that o be conserved. In h2, however, o is not conserved. So, if C2 is true, she has a pro tanto reason to disprefer h2 rather than h1 being the case—namely, that in h2 a presently existing valuable thing ceases to exist at t, whereas in h1, no valuable thing ever ceases to exist.
The claims above are about what Derek Parfit calls this person's telic reasons—the reasons given by the ways in which ‘the event that we want would be in itself good or worth achieving’ (2011: 50). This person has some such telic reasons to want h2's being the case and to want h1's being the case. My phrase ‘reason to disprefer h2 rather than h1 being the case’ is thus shorthand for ‘Considering the telic reasons to want h1 to be the case and the telic object-given reasons to want h2 to be the case, the latter set of reasons is less weighty than the former’. I am assuming here that all other facts are held fixed—that is, there are no other causal consequences of the person's preferring that h2 rather than h1 be the case. (Indeed, for readers willing to countenance normative reasons in the absence of any persons, the presence of this agent in the case is entirely optional.)
Now since C2 implies that the agent has a reason that no version of standard pluralist consequentialism implies that she has, the Mirrored Histories case meets the desideratum set out at the end of section 2. It trivially avoids the complications raised by cases like All Souls Enhancement, since there is no question of different kinds of value existing in each of its possible histories. To the extent that it provides intuitive support for C2, it therefore favors C2 over all version of standard pluralist consequentialism considered thus far.Footnote 2
Since it sharply distinguishes such standard pluralist consequentialist views from C2, the Mirrored Histories case also suggests the following argument for C2: If C2 is not true, there is no reason to disprefer h2 to h1; but, intuitively, there is such a reason; so, C2 is true. You will find this argument persuasive only if, like me and other Cohen-conservatives, you share the judgement that there is a reason to disprefer h2 to h1 (henceforth, for short, ‘the h2 < h1 judgment’). (The case might also be used to test the weight of the conservative reason, by progressively decreasing the quantity of value borne by o in h1 and locating the point at which one judges there to be no more reason to disprefer h2 to h1.)
To nonconservatives, the case will, of course, suggest a parallel modus tollens case against C2. A more complete defense of C2, then, would vindicate the judgment in question—that is, it would explain in virtue of what there is such a reason (or, at least, would show why there is no such informative explanation to be had). My aim is more modest. I am aiming merely to show that, to the extent that the Mirrored Histories case provides intuitive support for C2, that support is undefeated. Thus I next consider objections which, in different ways, question whether whatever intuitive support the Mirrored Histories case confers on C2 is undefeated.
Before turning to those objections, it is worth noting that the Mirrored Histories case would need fewer assumptions if C3, not C2, best reflected Cohen-conservatism. We could then drop the assumption that there is any moment of time at which o would exist in both of h1 and h2, and we could drop the assumption that the person in question is evaluating these histories at any particular time. Since o is independently existing—since it will (have) exist(ed), no matter what anyone does—we could simply take this person to be asking, either timelessly or retrospectively, whether there is any reason to disprefer h2 to h1 having eventuated. C3, unlike any version of standard pluralist consequentialism, would entail that there is such a reason.
4. How Well Does the Mirrored Histories Case Support C2?
I have argued that, to the extent that the h2 < h1 judgment confers intuitive support on C2, that support is undefeated with respect to a wide range of standard axiological views. In this section, I consider five objections that, in various ways, question whether the intuitive support that this judgment can confer on C2 is, ultimately, undefeated. Each objection claims that the h2 < h1 judgment can be vindicated, without appeal to C2, by recognizing some pattern fact about a history to be a basic source of (dis)value (or to be otherwise reason-implying). Pattern facts are, as Broome puts it, facts that ‘show up as a pattern contained in the distribution [of goods and bads in a history]’ (2004: 44). The objections differ as to what they take the relevant pattern fact to be.
4.1 Bias toward the Future
Parfit famously suggests that people exhibit a ‘bias toward the future [which] applies most clearly to events that are in themselves pleasant or painful. The thought of such events affects us more when they are in the future rather than the past. Looking forward to a pleasure is, in general, more pleasant than looking back upon it’ (1984: 160).
The bias would cause us, if we could, to postpone pleasures for the future and bring pains (even at the cost of making them greater) nearer, so as to get them behind us as soon as possible. The bias toward the future, one might object, explains the h2 < h1 judgment better than C2 does. The proposal is that the bias toward the future affects, not only judgments about our own good, but judgments about goodness simpliciter: we prefer that goods lie ahead of us, not behind us, in time. There is some reason to believe that such a bias is not, in the prudential case, irrational (Parfit Reference Parfit1984: 186; for discussion of Parfit's considered position on this question, see Scheffler Reference Scheffler, McMahan, Campbell, Goodrich and Ramakrishnan2021). And so, in the present context, this explanation of the h2 < h1 judgment could take either a vindicating form (suggesting that this preference is not irrational) or a non-vindicating form (suggesting that the preference is an instance of a bias that could, at most, be rational only in the prudential case).
I have three replies to this objection. First, it is questionable whether this explanation of the h2 < h1 judgment would succeed in its vindicating form. The bias toward the future seems most clearly rational in the case of mere pleasures and pains. As Parfit notes, it gives us no succor that a shameful event is in the past, and we feel no inclination to keep events that would be cause for pride in the future (Parfit Reference Parfit1984: 160). In considering its application to the Mirrored Histories case, we can ask whether the prospect that the future will contain a finally valuable object (controlling for any pleasure that we might take in it) is more like the prospect of a future pleasurable experience or more like an event that will be the occasion for justified pride. It seems reasonable to think it more like the latter category. Events that occasion pride or shame, as Parfit claims, ‘gild and stain’ (1984: 160) our lives. Their goodness or badness is in that sense persistent, not limited only to the moments at which we are experiencing them. That is why the bias toward the future does not gain purchase with respect to them. Given the simplifying assumption made in the final paragraph of section 1, the existence of finally valuable things seems relevantly similar. Their value, too, does not solely depend on our experiencing them; as one might put it, their existence gilds the history of the universe. So there is reason to doubt that the h2 < h1 judgment could be given a vindicating explanation by appeal to the bias towards the future.
Consider, second, the non-vindicating version of the objection. To test whether this error theory would undermine the intuitive support that the Mirrored Histories case confers on C2, we can consider a variant of the case. Thus suppose that, rather than overlapping for a moment at t, there would be a period of forty years during which o would certainly exist in both h1 and h2. And suppose that the person in question wakes up as this forty-year period, which roughly coincides with the expected remainder of her life, commences. Since this person would co-exist with o for the remainder of her life, no matter whether she is in h1 or h2, the h2 < h1 judgment could not plausibly be explained away on the grounds that we have a bias toward the future.
Third, and finally, the possibility of this alternative explanation arises only on the assumption that C2, not C3, is the best formulation of Cohen-conservatism. The alternative version of the Mirrored Histories case described at the end of section 3, which asks a retrospective or timeless question about how h1 and h2 compare, would not be vulnerable to this error theory. So, to the extent that C3 is a plausible reconstruction of Cohen-conservatism, it is questionable whether this error theory could explain away the h2 < h1 judgment.
4.2 The (Value) Shape of a History
J. David Velleman (Reference Velleman1991) suggests that the shape of a life partly determines its prudential value. Of two lives with the same total amount of momentary and lifetime welfare, a life with momentary welfare that is an increasing function in time is better than one in which momentary welfare is decreasing in time. The value of a life depends in part on this pattern fact—a fact about how a person's welfare is distributed over different times. One might propose that the value of a history can depend on a similar pattern fact—in particular, that there is a reason to disprefer a history to the extent that, in that history, total value in the universe is a decreasing function in time.Footnote 3 On this view, as one might put it, the ‘(value) shape of a history’ partly determines its overall value. But only in h2 is total value a decreasing function in time. So this view entails there is reason to disprefer h2 to h1 and, so it might be objected, better explains this judgment than C2.
I have two responses. First, it is not easy to see how to motivate the (value) shape of a history view. Although I noted an analogy between this view and the shape of a life view, the considerations that motivate the latter do not obviously lend support to the former. As Velleman suggests, the latter view would best be supported by appeal to the claim that the narrative of a life—roughly, the meaning that different events have within a life—matters. But while it is plausible that the narrative structure of a life matters to its prudential value, it is harder to see why the narrative structure of a history of the universe should, in its own right, matter to that history's value. A history is not a life, and there does not seem to be sense (or, at least, morally relevant sense) to be made of a history's having a perspective in which such a narrative could be anchored or grounded.
Suppose, however, that the (value) shape of a history view could be motivated. This would nevertheless not be enough to undermine the support that the Mirrored Histories case confers on C2. To see this, consider a variant of the case. This variant differs from the original case in three respects: (a) In addition to o, each possible history contains a second intrinsically valuable thing, p, which exists at all times and bears the same kind of value as o; (b) p is such that: in h1 (respectively: in h2), its total value until t is 2 (respectively: 1), and its total value in the time after t is 1 (respectively: 2); and (c) in both h1 and h2, the total value of o is 1. In both h1 and h2, then, the value of the universe is a constant function in time. But, I take it, conservatives will find the h2 < h1 judgment persistent in this case, when they focus on the fact that only in h2 does a valuable thing fail to be conserved. Since the (value) shape of a history view would not fully accommodate and explain conservatives’ judgments, it would not therefore defeat the intuitive support that the Mirrored Histories case confers on C2.
4.3 Wasted Possibility
It might next be objected that there is a different explanation of the h2 < h1 judgment: namely, that there is what one might call ‘wasted possibility’ in h2. The proposal is that, for most people, a particularly vivid or salient thought about h2 is that o could have gone on existing, but that the corresponding thought about h1 (that is, that o could have existed earlier than it exists) is much less vivid or salient (cp. Kagan Reference Kagan2012: 227). For this reason, the objector suggests, people are mistakenly inclined to judge that there is a reason to disprefer h2 to h1. This explanation of the h2 < h1 judgment might instead take a vindicating form. The proposal would be that it is better that valuable things not only exist but do so in a way that is modally robust—compare Felix Pinkert's (Reference Pinkert2015) suggestion that morality imposes a requirement not only to act rightly but to do so in a way that is modally robust—and that this seems to be true of o in h1 but not in h2.Footnote 4
In reply, consider another variant of the Mirrored Histories case, in which each possible history also contains an unbiased coin, which is flipped once a day. If the coin shows heads, o is either created (if it does not exist) or sustained in existence (if it already exists); if the coin shows tails, o is either destroyed (if it exists) or prevented from coming into existence (if it does not exist). In h1, there is an infinite series of tails until t, and then an infinite series of heads; in h2, there is an infinite series of heads until t, and then an infinite series of tails thereafter. The coin flips make vivid the fact that, if there is wasted possibility in either h1 or h2, there is wasted possibility in both. But, I take it, those who share the conservative intuitive judgment about the original Mirrored Histories case will again judge, on the same grounds as before, that there is a reason to disprefer h2 to h1. Thus, this explanation, in either its vindicating or non-vindicating form, does not defeat the intuitive support that the case confers on C2.
4.4 Welcoming the New
It might next be objected that there is a still further explanation of the h2 < h1 judgment—namely, that in h1, but not in h2, something valuable comes into existence. This view is, in a sense, C2's dual; I denote it by C*. According to C*, there is a pro tanto reason to welcome the coming of the new (as opposed to, what C2 implies, a reason to regret the passing of the old). If C* is true, then C2 is not required to explain the h2 < h1 judgment. (This is related to, but distinct from, Anca Gheaus's suggestion, reported in Cohen [2012: 165], that there is sometimes reason to welcome change for its own sake.)
In reply, consider another variant of the Mirrored Histories case. This variant differs from the original case in three respects:
1. Time extends infinitely only in the forward direction—that is, there is an earliest time, t – n, after which time continues without end;Footnote 5
2. In h1, as in the original case, o comes into existence at t and never ceases to exist. In each hn, o has always existed (that is, o comes into existence at t – n) and continues to exist until t; and
3. Given that o begins to exist at some t1, the value of o at tn is given by
$v_{t_n}\left( o \right) = \; \mathop \sum \limits_{i = 1}^n \displaystyle{1 \over {2^i}} $ with o's value, in the limit, being 1.
Figure 6 illustrates this variant case:
Figure 6. A variant of Mirrored Histories
Now consider one view that incorporates C2 and another that incorporates C*, and contrast their verdicts about this case. These two views agree that, for any finite n, h1 is in one respect better than hn. For any finite n, the amount of value of which o is the bearer is greater in h1 than in hn. However, according to the view that incorporates C*, for large n, the difference in value between h1 and hn becomes negligible (as the function that gives the value of o approaches its limit). On the other hand, according to the view that incorporates C2, there is an additional reason to disprefer hn to h1. (According to the view that incorporates C*, there is no corresponding respect in which any hn is better than h1, since, in both h1 and each hn, there is a time at which o begins to exist.) Thus, for large n, a view that incorporates only C* could not explain conservatives’ judgments about the weight of the reasons to prefer h1 to hn. So, even assuming C* is correct, it does not defeat the intuitive support that the Mirrored Histories case confers on C2.
4.5 Against Mere Destruction
I have so far considered four objections that proffer alternative explanations of the h2 < h1 judgment, and I have concluded that they do not undermine the intuitive support that the Mirrored Histories case confers on C2. The final objection, however, merits different treatment. It suggests a genuine competitor to conservatism that I believe cannot easily be dismissed.Footnote 6 According to this objection, as before, there is a different view that explains the h2 < h1 judgment. It is the view that there is reason to prevent (or regret) the destruction or alteration of a finally valuable thing, over and above the reason given by the value it bears. In the Mirrored Histories case, only in h2 is a valuable thing destroyed. So the view in question entails that there is a reason to disprefer h2 to h1.
This view may seem, at first glance, merely to redescribe C2. But it differs importantly from C2. One way to prevent the destruction of a valuable thing is to prevent that thing from ever existing. The view in question would claim that, if preventing some valuable thing from ever existing is (as it often in fact is) the only way to avoid its being destroyed, then there is a reason to prevent that thing from ever existing. (The view in question is therefore analogous to so-called anti-natalist views (Benatar Reference Benatar2008) about creating new persons. If creating a valuable thing would mean its later being destroyed or significantly altered, then this view says, ‘Better for it never to be.’) That is not what conservatives favor. Of those valuable things that there (presently or independently) are, conservatives take there to be a distinctive reason to prevent (or regret) their destruction. They call for holding on to the value we have. But they do not call for valuable things not to be created. Since the view in question, by contrast, is merely concerned with avoiding destruction (or significant alteration), call this the mere destruction view.
Here is another way to see the difference between C2 and the mere destruction view, using the representational device introduced in section 1. The mere destruction view condemns a certain pattern fact: histories in which there is a row with both a numerical symbol and then a dash. There are two ways of avoiding one's history having this pattern: by removing the numerical symbol altogether (that is, by causing an object never to exist) or by conserving the object once it exists. The mere destruction view entails that there are reasons to do either kind of action. C2 entails only that there is a reason to do the latter kind of action.
Both C2 and the mere destruction view would accommodate and explain the h2 < h1 judgment. So, while the h2 < h1 judgment provides favors C2 over a wide range of standard and nonstandard pluralist consequentialist views, as discussed above, it does not favor C2 over the mere destruction view. The conservative might respond that C2 is the only plausible view that explains the h2 < h1 judgment. To some, the mere destruction view may seem to express an attitude toward value that is repugnant if not incoherent (for reasons that mirror the reactions that some people have toward the analogous anti-natalist view). But there are at least two reasons to take the mere destruction view seriously. First, it is the analog of a view about (at least one aspect of) the badness of death that arguably has some plausibility. F. M. Kamm claims that ‘one's conscious life being all over . . . is a bad aspect of death independent of its depriving of us of additional goods’ (2017: 734, emphasis added). In support of this claim, Kamm suggests that we would judge it in one way better, if we could, to go into a limbo state for a long period of time, delaying our deaths while not having any additional goods (Kamm Reference Kamm2017: 734–35). The mere destruction view might be motivated by analogy with this ‘allover-ist’ view about (one aspect of) the badness of death.
The second reason that I believe the mere destruction view cannot be dismissed can be illustrated by contrasting the pattern of concern characteristic of C2 with that of the mere destruction view. Focus on some time slice t. The mere destruction view, as noted above, has a simple pattern of concern. Whether or not some valuable thing o yet exists at t or not, the view entails that there is always a reason disfavoring its destruction. But given that o will (or given that it will not) be destroyed at t+n, the mere destruction view is indifferent between its yet existing at t or not. Figure 7 illustrates the view's pattern of concern.
Figure 7. The mere destruction view's pattern of concern
That simple, unconditional pattern of concern is unproblematic. Consider, by contrast, C2's pattern of concern. If o exists at t, C2 entails that there is a reason disfavoring its destruction. But, if o does not exist at t, then C2 is indifferent between its later destruction or its not being destroyed. And, crucially, C2 does not express a view about whether a valuable thing should exist at t or not. It displays a conditional preference for there not to be destruction of a valuable thing, on the condition that it exists, but it is, in Broome's phrase, ‘neutral about the condition [itself]’ (2004: 154). If we encode this neutrality about the condition as reflecting equal choiceworthiness or preferability, then C2 has the pattern of concern shown in figure 8.
Figure 8. C2's pattern of concern
But, as Broome (Reference Broome2004: 156–57) argues, views that display this pattern of concern are quite unlike everyday notions of conditional betterness or conditional choiceworthiness. They display a non-transitive pattern of concern.
I do not claim that the foregoing considerations constitute decisive reason to reject C2. First, when Broome objects to such patterns of concern, he is objecting to one kind of view about (conditional) betterness. But C2 is not a view about the value of outcomes. It is arguably true that, by virtue of the meaning of ‘betterness or the role it plays in practical reasoning, the betterness relation could not display non-transitivity (Broome Reference Broome2004: ch. 4; for dissent, see Temkin Reference Temkin2012). But it is less obvious that failures of transitivity in the domain of normative reasons are similarly objectionable (Broome Reference Broome2004: 60). Second, and relatedly, the assumption that ‘neutrality about the condition’ should be encoded as equal choiceworthiness is questionable. The more modest point I wish to make is that, since the mere destruction view vindicates the h2 < h1 judgment, on a basis that arguably can be independently motivated, and without requiring us to avoid, or explain away, the threat of non-transitivity in our pattern of concern, there seems to be (defeasible) reason to prefer it. It may be that there is, on other grounds, a decisive case for rejecting the view. If so, then I conclude that whatever intuitive support that the Mirrored Histories case confers on C2 is undefeated. If not, then the Mirrored Histories case would have the merit of showing that C2—even if it better explains a set of intuitive case judgments than a wide range of standard and nonstandard versions of pluralist consequentialism—would need further defense before one could conclude that it best explained those judgments.
4.6 Is the Mirrored Histories Case Faithful to Underlying Conservative Attitudes?
Some people who have conservative judgments may balk at the idea of appealing to the abstract and unrealistic Mirrored Histories case in support of their view. Making such an appeal may seem to undermine the attitude toward value and valuable things that they take their view to express. That attitude is rooted in a distinctive kind of attachment to actually existing things, in all their rich particularity. The Mirrored Histories case, by contrast—and by contrast with the cases to which Cohen appeals—calls for making evaluative judgments about a thinly described object, that is merely stipulated to be valuable, in a highly unusual choice context. That may seem objectionably untethered from our actual valuing attitudes and practices concerning the valuable things that there actually are.Footnote 7
It may indeed be that some people accept C2 on the basis of a view about value and valuing that would undermine an appeal to the Mirrored Histories case. My arguments therefore raise a challenge for these people: to investigate whether there are any alternative sources of undefeated intuitive support for their view. But to accept C2 need not be to presuppose such any such view of value or valuing. As Cohen (Reference Cohen and Otsuka2012: 167) notes, although one aspect of conservative attitudes may essentially depend on the presence of certain kinds of historical or personal connections between valuer and object (the aspect Cohen calls ‘personal valuing’), the ‘particular valuing’ attitude reflected in C2 does not. So it is no more out of place to consider a case like Mirrored Histories to test the plausibility of C2 than it would be to consider a hypothetical utility-monster case in order to test the plausibility of utilitarianism.
One might alternatively object that, insofar as C2 is no more than a ‘conditionalized’ version of the mere destruction view I discuss in section 4.5, it differs in spirit either from the view discussed by Cohen or from any distinctively conservative view. That it is a close cousin of the mere destruction view might be thought to suggest that C2 still, in effect, exhibits only a concern for an impersonal bad—namely, the mere destruction or alteration of (a subset of) what is valuable—rather than a concern that is tied, in some more robust sense, to particular bearers of value. C2 would, for example, favor the destruction of one presently existing valuable thing for the sake of preventing the destruction of five presently existing valuable things (Cohen Reference Cohen and Otsuka2012: 156). And it might therefore be claimed to be unfaithful to a nonconsequentialist strand in the attitude that underlies conservatism—a strand that, in turn, reflects a concern that is essentially tied to the valuing of particulars.
I agree that one could develop a version of Cohen-conservatism that, in the case referenced in the previous paragraph, would not favor the destruction of one presently existing valuable thing, even to prevent the destruction of five others. (This version of the view would, in effect, take conservatism to give people a goal that is to be respected, not promoted.) But I deny that C2's implications about this kind of case make it unfaithful to the motivations underlying conservatism. Accepting C2 does not entail accepting that mere destruction is impersonally bad; indeed, that claim is more difficult to reconcile with C2 than with the mere destruction view. And claiming that the C2-given reason corresponds to a value to be promoted (or, more precisely, a disvalue that is to be dis-promoted) would not undermine the view's conservative credentials (as noted by Cohen Reference Cohen and Otsuka2012: 156).
I conclude that, if one shares Cohen's intuitive judgments, then (modulo the complication raised by the mere destruction view) the Mirrored Histories case confers undefeated intuitive support on C2. Of course, it may well be that some people, although they share the h2 < h1 judgment in the original case, lose the corresponding judgment in some of the variant cases discussed in subsections 4.1–4.5. I take this to be a further merit of the Mirrored Histories case: it allows more careful reflection on whether one's intuitive judgments really do support C2. (Recall that my aim has not been to defend C2; it has been, rather, to isolate a kind of case from which, if the view could draw intuitive support, it could draw undefeated support.) Even for those who do share these judgments, however, the mere destruction view may raise a further challenge for defenders of conservatism. It would arguably give them a further explanatory burden to discharge—namely, to explain what grounds the putative reason to regret the destruction of only presently or independently existing valuable things. They would face this challenge if it is true that, when two views equally well accommodate the intuitive data, there is more reason to accept the view that can better explain and defend the grounds of that judgment (Bengson, Cuneo, and Shafer-Landau Reference Bengson, Cuneo and Shafer-Landau2019). The Mirrored Histories case also casts some doubt on other reconstructions of Cohen's view in the literature. Ralf Bader (Reference Bader, Cohen and Otsuka2013) suggests that Cohen is best understood as assigning, to each valuable thing, a further kind of ‘particular value’. But this view would not recover the h2 < h1 judgment, since there would be equal quantities of this value in both histories. Geoffrey Brennan and Alan Hamlin (Reference Brennan and Hamlin2016b) suggest that Cohen's view is best understood as positing a ‘state-relative’ reason—a reason to conserve independently existing things only given that some state of affairs is actual. But suppose that, in the variant Mirrored Histories case presented at the end of section 3, God is deciding whether to make h1 or h2 actual (and has no other options). This view would fail to recover the result that God would have conservative reason to make h1 actual.
5. Conclusion
There is reason to question whether Cohen's conservative thesis, whether reconstructed as C2 or C3, can draw undefeated intuitive support from the cases to which Cohen appeals. But, to the extent that one shares Cohen's judgments, there is a Mirrored Histories case that favors Cohen's conservative thesis over a wide range of standard pluralist consequentialist views—even those that recognize fine-grained distinctions or interaction effects among the basic sources of value. Conservatives can reasonably be expected to share Cohen-type judgments about this case. So the case allows them to claim undefeated intuitive support for their view, with respect to a wide range of alternative axiological views. Reflecting on the Mirrored Histories case is also illuminating in a further way. It suggests a different view, one that sets itself in opposition to the mere destruction or alteration of finally valuable things, which would also vindicate many apparently conservative judgments. That view arguably merits consideration in its own right.
