Abstract
Some theorists believe that there is a plurality of values, and that in many circumstances these values are incommensurable, or at least incomparable. Others believe that all values are reducible to a single super-value, or that even if there is a plurality of irreducible values these values are commensurable. But I will argue that both sides have got it wrong. Values are neither commensurable nor incommensurable, at least not in the way most people think. We are free to believe in incommensurability or not, depending on what particular conception of morality we want to embrace. Incommensurability is accordingly not a theory about value. It is a presupposition that provides a necessary background condition for a certain kind of value to exist. It is therefore not the kind of view that can be morally true or false. As a presupposition, it can only be accepted or rejected on grounds that do not presuppose that morality already exists. Incommensurability is, like the rejection of hard determinism, one of the presuppositions on which morality as we know it happens to be based.
Keywords
For purposes of this paper, I will take the value of any two goods, acts, omissions, states of affairs, or even the values of values themselves to be incommensurable if the value of one is neither more than, less than, nor equal to the value to the other. Some theorists reserve this definition for incomparability—which they distinguish from incommensurability, arguing that commensurability is more demanding, for it requires that value be measurable according to a common scale, while comparability merely requires that ordinal rankings be possible, and this could be the case even though cardinal ones are not (see, for example, Chang, 1997: 2). But I reject this distinction. I think that values cannot be incommensurable but comparable, and that much confusion has been caused by attempts to suggest otherwise. I will defend this claim at length later in this paper, but for now the distinction between incommensurability and incomparability is not important. For now, those who feel more comfortable with the term “incomparability” to describe situations in which the value of one option, good, act, state of affairs, or whatever is neither more than, less than, nor equal in value to another should feel free to substitute that term wherever the term “incommensurability” appears.
Some theorists believe that there is a plurality of values, and that in many circumstances these values (and therefore the value of the options, goods, acts and states of affairs that contain them) are incommensurable. Others believe that all values are reducible to a single super-value, or that even if there is a plurality of irreducible values these values are commensurable. Regardless of which view they advocate, however, everyone seems to agree that those on one side of the controversy are right, while those on the other side are wrong, even if it is not clear yet which side is which. 1 The continuing failure of one side to convince the other has no more significance than the failure of everyone to agree on what morality requires. We simply have not got our thinking on these matters clear enough yet. Once we are able to identify what is leading some of us into error, all this disagreement will disappear. Everyone will see that values are commensurable, or that they are not—whichever view turns out to be correct.
But once again, I do not think this is correct. It is tautologically true, of course, that either values are incommensurable or it is not the case that values are incommensurable, but it is not true that determining whether values are commensurable is simply a matter of properly interpreting whatever moral theory we happen to embrace, and that one side or other here is simply committing a mistake. Deciding whether values are commensurable or incommensurable is not something we can do after deciding what conception of morality to embrace. On the contrary, it is something we must do before embracing a particular conception of morality, or at least before embracing a certain kind of conception of morality. Indeed, what I shall argue in the first part of this paper is that a belief in incommensurability is not a belief that can itself be morally true or false under any particular moral theory—rather, it is a presupposition that is necessary for a certain kind of moral theory to be able to exist. Instead of being determined by our moral theory, it is something that determines the range of moral theories that are available to us to accept. As a moral presupposition, it sets the terms of moral debate, and determines what kinds of arguments are permissible to make.
In this sense, incommensurability is to morality what H. L. A. Hart’s rule of recognition (1994: 100–110) is to law. Just as the rule of recognition is the presupposition that determines what “law” is and therefore cannot itself be legally true or false but merely accepted or rejected, it can be right or wrong to accept or reject incommensurability only in the pre-moral sense that there may be good reasons to go one way rather than the other before we have decided what our moral values are. I shall not have much to say about what these reasons might be, except to note that we seem to find the concept of value, and especially moral value, very useful as a means of regulating social interaction and cooperation. 2 There may be other reasons that apply here too, but for the purposes of this paper I shall simply assume that the utility of morality is a sufficient reason for its existence, and argue that morality could not have such utility if we did not accept incommensurability. Incommensurability, like the rejection of hard determinism—the view that humans have no free will and therefore cannot be morally responsible for their actions—is one of the presuppositions on which the conception of morality as we know it—the conception of morality that most of us embrace—is based.
In the second part of the paper, I take on the claim that values can be comparable even when they are incommensurable. I take this claim to be the most powerful potential objection to my argument, for comparability is all that is required for morality as we know it to exist, so if this claim is true it renders my opening argument moot. What I shall argue, however, is that values can be incommensurable and incomparable, or commensurable and comparable, but they cannot be one without the other. If I am right about this, my opening argument survives, and while there might be conceptions of morality that do not presuppose incommensurability, these are not conceptions that many of us accept. On the contrary, for morality as we know it to exist, we must believe that values are incommensurable, or at least that widespread, pervasive, and important incommensurabilities exist.
In the third part of the paper, I will sketch out a resolution to the apparent paradox that the first two sections of my argument seem to create: the idea that while incommensurability is necessary for morality as we know it to exist, incommensurability also renders morality useless as a guide for human conduct in precisely the situations in which we presumably need it most—situations in which we must choose to pursue one important value or another, but cannot pursue both. That this seems to present a paradox explains why the debate over incommensurability has proved so intractable for so many years. Many people reject incommensurability because they fear that doing otherwise will leave insufficient territory for morality to operate, for morality must be able to help us make difficult choices between hard-to-compare actions and/or states of affairs if it is to have the utility that justifies accepting the presuppositions that allow it to exist. 3 To alleviate these concerns, we must show that moral reasoning can proceed even when comparisons between carriers of value are not possible. This, in turn, will resolve the paradox that an incommensurability-based conception of morality seemingly creates, and therefore make the argument for incommensurability much harder to resist.
Before I move on to the substance of my argument, however, I want to point out something about the method I will employ. Most arguments for or against incommensurability are what we might call micro-arguments—they proceed by way of example and counter-example. One side offers choice situations in which the available options seem incommensurable; the other tries to show that these options really are commensurable, or at least comparable. In both cases, each side tries to build their argument from the bottom up, by arguing that accepting the view they promote is the only way to explain our largely intuitive reactions to these examples. 4 In contrast, the argument I will offer is a macro-argument. Instead of starting with a series of examples and insisting that accepting incommensurability is the only way to explain our intuitive reactions to them, I will employ a conceptual argument that applies no matter what choices are at issue. Instead of arguing from the bottom up, I will be arguing from the top down.
Exactly what this means should become clear in a moment, but I do want to point out at the outset that given the method of my argument, it cannot be proved wrong by counter-example. At least it cannot be proved wrong by counter-example alone. Counter-examples may be useful for illustrating that our intuitions are not fully aligned with a conceptual explanation, but only an explanation of where the conceptual argument has gone wrong can actually defeat it. Without such an explanation, appealing counter-examples could just signal that our intuitions are wrong, or that our conceptual argument is currently incomplete, rather than incorrect.
With this in mind, we can now move on to the first stage of my argument: the claim that a belief in incommensurability is one of the circumstances of justice—that is, one of the presuppositions necessary for morality as we know it to exist.
1. Incommensurability and the circumstances of justice
According to the well-known observation by Hume, there are certain background circumstances that must be present in order for questions of justice to arise. For one thing, people must tend to be more concerned about their own well-being than they are about the well-being of others. For another, the combined amount of resources to be found in nature or produced by social cooperation must be insufficient to satisfy everyone’s wants and desires. Hume mentions various other circumstances as well, but he focuses primarily on these two—limited altruism and limited resources—and argues that as long as these circumstances characterize the world in which we find ourselves, disputes over how available resources should be divided will be endemic in our social life. In order to resolve such disputes, we accordingly need principles of justice to tell us what patterns of distribution morality permits, prohibits, and requires. If we lived in a world of either perfect altruism or true superabundance, in contrast, principles of justice would be superfluous, for disputes about how the available resources should be distributed would in this case never actually arise. 5
Hume’s reflections on the circumstances of justice and the assumption that these circumstances apply underlie every discussion of distributive justice I can think of, and it is widely thought that Hume’s reflections on these matters are correct. 6 Of course, Hume may have been referring only to the circumstances of distributive justice—the rules that regulate the distribution of the benefits and burdens of social cooperation, and specifically the rules of property and of contract. 7 But Hume’s remarks could also be understood to refer to all the rules of morality, not merely the rules of property and contract, and that is how I am going to interpret him here. I am going to take Hume’s claim regarding the circumstances of justice to be a claim about what conditions are necessary and sufficient for the need for any principles of morality to arise. 8
Should you disagree with this interpretation, however, do not worry—there is no reason for concern. I will be using Hume’s discussion of the circumstances of justice as a launching point for a series of hypothetical thought experiments. These thought experiments are designed to extend Hume’s thinking across ever more extreme situations, and they therefore intentionally go beyond what Hume had in mind. So even if Hume is better read in a more limited fashion, this will merely adjust the point at which my extensions of Hume are deemed to begin. It will not affect the relevance of these extensions, nor the importance or soundness of any of the conclusions that I base thereon.
Is it true, then, that absent the circumstances of justice there would be no need for interpersonal moral principles of any kind? It is my contention that it is not. Neither limited altruism nor limited resources are individually, jointly, or alternatively necessary conditions for questions of justice to arise. Nor are they individually, jointly, or alternatively sufficient conditions. The amount of resources available for human use and consumption and the degree of altruism characteristic of human nature has an effect on the substance of the principles of justice required, but neither factor is relevant to determining whether some set of principles are required, full stop. With regard to limited altruism, however, I shall not attempt to defend my contention, for nothing I want to say about incommensurability turns on this. 9 My contention regarding limited resources is another matter. We would need moral principles to regulate our conduct even if resources were unlimited for, as we shall see, unlimited resources and full commensurability are related concepts. And there are some important insights about incommensurability and the role that a belief in incommensurability plays in the structure of most people’s moral reasoning to be gained from recognizing this.
1.1. Limited resources and the circumstances of justice
To explore the extent to which resources must be limited for the need for principles of justice to arise, we must first address how the absence of limited resources—call this superabundance—is to be defined. If, for example, superabundance applies only to social primary goods—rights, liberties, and opportunities, wealth and income, and the bases for self-respect—questions of justice could still arise, for deficiencies in natural primary goods, such as health, vigor, intelligence, imagination, and natural talents and abilities, as well as in the ability to convert resources into welfare, might lead to calls by some for compensation or even leveling down. Interference with such goods would certainly give rise to claims for compensation, and perhaps even punishment and retribution—claims that would require moral principles to adjudicate, at least as long as natural primary goods were not themselves in unlimited supply. Most importantly, even with a superabundance of social primary goods, we would still have reason to form emotional attachments to others, to value friendship, family, respect, knowledge, achievement, beauty, and also time. So interference with the sources of such value, and perhaps even natural deficiencies in or naturally occurring damage to them, would also inflict injustice (at least in some cases) and give rise to legitimate claims for punishment, compensation, and/or repair.
Suppose, for example, that your wife is killed in a traffic accident caused by the negligence of another. Unless it is possible to bring her back to life, or to replace her with someone with the same attributes and memories and attitudes, you (and of course she as well) would experience a loss, a loss that required principles of justice to redress. If superabundance is to relieve us of the need for principles of justice, it must accordingly include much more than an unlimited supply of social primary goods. To relieve us of the need for principles of justice of every kind, superabundance must be understood as meaning that everything is replaceable, including not only natural as well as social primary goods, but also specific people and parts of people and even time. There must be an unlimited supply of everything that anyone might value.
If we were to define superabundance in this way, then nothing would be unique. Every carrier of value would be part of an infinitely large set, and while I might have reason to form attachments to that set, I would have no reason to form attachments to any particular member of that set. If one member of that set were damaged or destroyed, it could be costlessly replaced. Human agents would be unable to deprive others of access to any carrier of value and therefore we would not need principles of justice to prohibit them from doing so or for providing remedies to those who might otherwise be injured when they do.
Even if our conception of superabundance were as broad as this, however, we would still need principles of justice for other purposes. While there would exist sufficient examples of every carrier of value for everyone’s preferences to be satisfied, we could still come into conflict. First, there would still be differences in our ability to convert resources into welfare. Those who were not able to achieve as high a level of welfare as others despite an unlimited supply of resources might press claims for leveling down. 10 We would therefore require principles of justice to adjudicate such claims and to devise appropriate remedies if the validity of such claims were ultimately established.
Second, while everyone would have sufficient resources to satisfy their personal preferences, their external preferences might still conflict. If I preferred that you suffered in abject poverty, for example, while you preferred to live in luxury, both our preference sets could not be satisfied even under conditions of superabundance. We would therefore need principles of justice to determine whether external preferences should be satisfied and, if so, how these were to be balanced against other preferences—external or personal—with which they happened to conflict. 11
Finally, while everything, including people and parts of people, could be costlessly replaced, sentient beings could still experience fear, pain, and other kinds of emotional and psychological injury. Because these experiences would be unpleasant, and could not be “un-experienced” even under conditions of superabundance, they would be experiences that such beings would desire to avoid. Such injuries would accordingly be morally significant, and we would need principles of justice to regulate their infliction and to set the terms of recompense whenever infliction was impermissible but nevertheless occurred.
To illustrate this last point, consider the film The Prestige. The film is about a magician who hires Nikola Tesla, the eccentric turn-of-the-century inventor and competitor of Thomas Edison’s, to build a machine—a teletransporter—that will allow the magician to perform a trick called “The Transported Man.” Tesla’s machine takes the form of a booth that the magician places on the center of the stage, and the instant the magician enters it he miraculously reappears at the back of the auditorium, to the audience’s amazement and delight. But the machine is not actually transporting the magician—it is duplicating him (the original remains in the booth on stage and it is the duplicate who appears at the back of the auditorium). This presents the magician with a problem. There are now two of him, and each time he performs the trick another will be created. To avoid the complications caused by this, the magician arranges for the one who enters the machine to fall through a trap door in the floor into a tank of water, where he will drown. The surviving duplicate will then become the new original, who will himself drown and be replaced by a new duplicate the next time the trick is performed. 12
Sorting out how we feel about the morality of these events is difficult and complex. Because one version of the magician dies just as his duplicate is created, a duplicate with all the same memories, attitudes, and characteristics as the original, there is a sense in which there is perfect continuity between the original and the duplicate, and therefore it might seem odd to think of the original as having been killed, much less wrongfully killed. And surely drowning himself as part of the trick is not the same as drowning someone else. It is not even clear that what happens here constitutes suicide, as it does not end with the magician ceasing to exist. While I think there clearly is something morally troubling about this (after all, the magician has deprived one branch of his post-teletransportation life-line of an opportunity to exist simply to maximize the convenience and entertainment value of a trick), those who disagree should focus on what the magician does when he first tests the machine. The magician does not arrange for his own branch life-line to be terminated. Instead, he places a revolver on the sideboard next to him, and when his duplicate is created and appears on the other side of the room, the magician shoots and kills him. In this case, the version of the magician that has been killed has not consented to be treated in this way, for even if we were to infer some sort of consent from the pre-teletransportation identity of the original and the duplicate, once the duplicate is created he possesses his own future—what Don Marquis (1989: 189–194) calls “a future like ours”—and intentionally putting an end to someone else’s future is morally significant. Before the duplicate can be put to death he must at the very least be given an opportunity to object, and he is not. It is accordingly hard to see how this act can be viewed as anything but murder. Even though there is an infinite set of magicians and each individual member of that set can be fully duplicated and therefore replaced (and, remember, if the magician can be duplicated, so can all his possessions, so each new magician could have the same possessions as the last—even the audience for each magician could be endlessly duplicated so each magician would still have the same fan base), the way in which an individual member of that potentially limitless set is treated can still constitute a moral wrong.
What this means is that the presence of limited resources is not a necessary condition for questions of justice to arise. Nor is it a sufficient condition. The extent of resources available in a particular society simply has no bearing whatsoever on whether that society requires principles of justice to regulate the interactions of its members. While any society which enjoyed unlimited resources would be a very different place than the society in which we actually live, even under the broadest possible conception of superabundance—one in which even people were totally replaceable—a need for moral principles would still exist. These principles might be more limited than the principles that seem to be required now, but they might not be more limited at all, and some principles of justice would be required no matter what.
1.2. Full commensurability and superabundance
Whatever problems would remain and require principles of justice to adjudicate in a world of superabundance, however, would be eliminated if all carriers of value were fully commensurable. Indeed, at least in a world of mere superabundance, different archetypes of values might be incommensurable even if individual instantiations of these archetypes were not—but if all carriers of value were fully commensurable, it would not matter if there were only a limited amount of cars, or Rembrandts, or children, or if people were unique in their attitudes, characteristics, and memories. We would no longer be concerned with categories, for there would be only one giant category—everything that one could value in any way. If everything were fully commensurable, uniqueness could not be a source of value; therefore the exchange of one sort of value for another could not be a morally significant event. People who were inefficient converters of resources into welfare could be compensated to bring them up to the level of welfare of everyone else, so there would be no need to consider leveling down. Those who harbored unsatisfied external preferences could be fully compensated, too, as could those who suffered any kind of experiential harm. Even those who were, as in The Prestige, deprived of “a future like ours” could be compensated in some way. So while superabundance would not eliminate the need for principles of justice, superabundance plus full commensurability would do so, for this would eliminate all the residual moral concerns that might arise under conditions of superabundance alone. 13
To see this, consider the reasons for action that would be available to an agent facing a choice between the following options: (a) do not commit a moral wrong; (b) commit the wrong but fully compensate those injured by it; or (c) commit the wrong but do not compensate anybody. In a world where all carriers of value are fully commensurable and there is superabundance, (b) must be possible, for whatever package of carriers of value were damaged by the wrong, there would be some other package of equal value that could be transferred from the agent who commits the wrong to those injured by it. Once that compensatory package was transferred, those injured by the wrong would have to be indifferent between states of affairs (a) and (b), for this is what being fully compensated means. 14 Even when that injury was death, the victim could be fully compensated, for given the assumptions we are operating under, the victim could be duplicated and replaced, and there would be some package of value that could be transferred to his duplicate that would make him (and therefore his original too, since each duplicate is identical to his original) indifferent to whatever harm had been originally inflicted. If those injured by the wrong were indifferent between these two states of affairs, then potential wrongdoers would have no reason to choose (a) over (b). 15
After hearing the above argument, however, some people have claimed that I am begging the question by defining choice (a) as committing a moral wrong, for when I do that, I am in effect assuming that there should be some way of distinguishing between (a) and (b), which is precisely the issue in dispute. If we do not assume that conduct (a) is a moral wrong, then there is nothing implausible about being morally unable to distinguish between (a) and (b). But that is precisely my point. In some cases, we think that causing harm to another is a moral wrong even if we do fully compensate those who experience that harm (most people would think this about torture, for example), and in some cases we think that causing harm to another is not a moral wrong even if we do not compensate those who experience that harm at all (most people would think that harming another by competing with them in business is not a moral wrong, for example, unless certain lines are crossed). I know of no case, however, where causing harm (defined here as causing a negative change in well-being to another) is wrong under morality as we know it only if we do not fully compensate those harmed.
Some people, however, have suggested to me that removing goods from a shop or eating in a restaurant is an example of a harm that is not a wrong as long as compensation is paid. But this is not the case. There is no harm here unless one intends to remove the goods or consume the meal without paying—not paying does not change an initial rightful act into a wrongful one; rather it allows us to determine that conduct which was initially morally ambiguous was actually wrongful. That is why belatedly paying the tab if one is caught trying to escape does not make the wrong of initially intending not to pay go away. It is still a wrong, or at least most people would think so, even if the full price of the good or meal is ultimately paid. If the failure to pay was not intentional, but merely negligent, then the failure to pay may be wrongful, but that does not change the character of the original act and make what was originally not wrongful suddenly become wrongful, or again most people would think so.
But there are other possible problem cases to consider here as well. A few people, for example, think that acts or omissions that trigger strict liability under tort law are examples of legal wrongs that are moral wrongs only if compensation is not paid, for no proof of traditional “fault” is required to establish legal liability, or at least they think that some people might think this (see, for example, Keeton 1959: 424). But I do not see how this view could possibly be correct. First, legal violations that trigger strict liability typically require that the activity or product in question be “abnormally” or “unreasonably” “dangerous.” 16 While a finding that an activity or product is abnormally or unreasonably dangerous is different than a finding of traditional fault, it is not that different. Strict liability violations accordingly do require a kind of fault in order for compensation for harm caused to be due, and therefore such violations do amount to a kind of wrongdoing even if whatever compensation becomes due is paid. Indeed, “when a court applies all of the factors suggested [by the abnormally or unreasonably dangerous standard] it is doing virtually the same thing as is done with the negligence concept” (Prosser et al. 1984: § 78, 555). Second, if legal violations that trigger strict liability are interpreted as not requiring fault of any kind, then it is unclear why the failure to pay compensation if it becomes due should be understood as turning the underlying legal violation into a moral wrong. Certainly no fault is required to prove that failure to pay compensation if it becomes due is itself a legal wrong, so why would we consider this failure to be a moral wrong, much less turn the underlying violation into a moral wrong without proof of fault if we did not consider the underlying violation to be a moral wrong in absence of proof of fault in the first place? Third, the more plausible description of these strict liability cases is that the initial conduct at issue here is only a moral wrong if it causes harm—and if it causes harm it is a wrong regardless of whether compensation is paid, although not paying may constitute a second wrong or make the total wrongdoing arising out of the incident more egregious. In other words, the wrongfulness of the initial conduct in these cases is indeed contingent, but it is contingent on harm being caused, not on payment of compensation. Finally, even if the only element that need be proved to make out a strict liability violation is causation, causation is at least partially a moral doctrine, not purely a factual one, and it is a moral doctrine in the corrective and not merely the distributive sense. Indeed, even those who claim that strict liability violations are wrongs only if compensation is not paid recognize this: “The notions of causation and fault are close kin. Picking one or more responsible causes from the multitude of antecedents of a given incident is very close to finding fault”(Keeton 1959: 402). 17
These same arguments also apply to “lesser evil” cases, the paradigmatic example of which (given by Joel Feinberg) is the day hiker who becomes caught on a mountain trail in a fierce snowstorm and breaks into an unoccupied mountain cabin in order to obtain shelter and prevent his otherwise certain death (Feinberg 1977). Judith Thompson, for one, thinks this kind of trespass to property rights is not wrongful, but the actor nevertheless owes compensation for what he has done. To explain this, Thomson (1977: 45–60) distinguishes between infringements of rights and violations, and suggests that infringements are not wrongful but do require the payment of compensation. Only if compensation is not paid in these cases does the trespasser’s conduct amount to a wrongful violation. But Thomson (1990: 164–165) concedes that what makes [the trespass of property rights] permissible or impermissible is an extremely complex affair, turning not only on the stringency of the [right], and the size of the increment of good to be got by infringing it, but on other things as well.
And she never makes clear how compensation could be required if infringement is not wrongful. Her argument actually seems to be driven by the concern that if the conduct at issue here were considered wrongful, it could be enjoined ex ante, yet it strikes her as morally inappropriate to allow this. Not all remedies, however, have to be available for the violation of all rights for those rights to nevertheless exist (see Reiff 2005 for an extensive discussion of this). Injunctions especially have never been granted as a matter of right, no matter how wrongful and damaging the threatened conduct is certain to be, and the burden is always on the party seeking the injunction to show that the balance of equities favors issuing it. Among other things, this means that the courts must be convinced that the threatened harm to the property owner is on balance greater if the injunction is not granted than the potential harm to the violator if it is (See Restatement (Second) of Torts (1979: § 936)). Because these cases usually involve minor property damage on the one hand and serious physical injury on the other, this will rarely be the case, so there is nothing problematic about treating the infringement of rights in these case as wrongful, for this does not commit us to the view that injunctive relief is an appropriate remedy. Even in Vincent v. Lake Erie Transp. Co., 109 Minn. 456, 124 N.W. 221 (1910), the case most often cited as an example of private necessity (the technical name for the “lesser evil” defense), the balance of the equities would not have favored injunctive relief, for the damage caused by defendant’s ship when it was tied up to plaintiff’s dock during a violent storm was far less than the damage the ship would have suffered if it had been forced to founder at sea. Thomson’s “solution” is accordingly a solution to a problem that does not exist. The actions of the real ship owner and the hypothetical hiker may not have been unreasonable, but they were clearly intentional and that is all that is required to establish fault. In other words, intentionally causing harm in these circumstances is indeed a wrong even though it is a lesser wrong than preventing someone from doing this; we need not come up with some other more convoluted explanation for this, and whether or not compensation is actually paid has no bearing whatsoever on the wrongfulness of the original act. In any event, in the real world, the very special facts necessary to make out a compelling lesser evil case almost never arise—in most cases, we have no trouble considering the conduct at issue prima facie wrongful even if there might be some extenuating circumstances explaining why the conduct at issue has occurred. 18
In any event, if there are no harms that are moral wrongs if and only if full compensation is not paid or, less categorically, even if only most harms that are moral wrongs are still wrongs even if full compensation is paid, then it is not begging the question to refer to moral wrongs rather than harms in the relevant hypothetical choice situation I have posed. But even if it were and we therefore had to replace the term “moral wrong” with “harm” in the various choice situations I have set forth above, we would still need to be able to morally distinguish (a) from (b), for we would still have to have a way of distinguishing harms that cannot be morally inflicted no matter what from harms that can be morally inflicted if fully compensated. And there is simply no way that we can do this without accepting incommensurability. So once again, only by accepting incommensurability can morality as we know it continue to exist.
But wait, these objectors say, there may be no way for consequentialists to believe in morality as we know it without accepting incommensurability, but those applying deontological theories of moral reasoning do this every day. Deontological theories put the right before the good, and therefore do not require us to distinguish between states of affairs to determine what morality requires us to do. Indeed, the whole point of deontological moral reasoning is to focus our attention not on states of affairs but on the nature of the acts and omissions involved, and if we look at the nature of the acts and omissions involved rather than at states of affairs, we can explain why some acts and omissions that cause harm are wrongs even if that harm is fully compensated.
Unfortunately, this is not correct; or rather it is correct only if we assume that we can distinguish between acts and omissions that are morally right or wrong completely independently of their actual, expected, or possible effects. I suppose there are a few deontological theories that purport to be able to do this—Kant’s categorical imperative would be one prominent example—but even most Kantians concede that the idea of strict adherence to the categorical imperative no matter how bad the consequences is not very appealing. Most of those who advocate deontological moral reasoning—including, I think it is fair to say, most contemporary Kantians—agree with John Rawls (1999: 26), who notes that Deontological theories are defined as non-teleological ones, not as ones that characterize the rightness of institutions and acts independently from their consequences. All ethical doctrines worth our consideration take consequences into account in judging rightness. One which did not would simply be irrational, crazy.
Indeed, it is almost impossible to think of any wrongful act that can be described without also describing its effects, at least in part. How would we describe murder, for example, without describing both the wrongful act (say, shooting someone) and its consequence (that is, death)? Even those applying deontological moral theories must accordingly consider states of affairs, or the consequences of acts, and not just how they came about, in order to distinguish between rightness and wrongness. 19 And if the consequences of two sets of acts are identical in all material respects, even considering how they came about—if two alternative sets of acts (on the one hand, do not commit a harm; on the other, commit the harm but fully compensate everyone harmed by it) will produce exactly the same effect, which is what would happen under the conditions of superabundance and full commensurability that we have postulated here, then it is difficult to see how even a deontologist could explain why it was wrong to engage in one set of acts but not the other. Only if we believe in incommensurability can we do this. Only then can potential wrongdoers have a reason to choose (a) over (b), and only then can we have morality as we know it. So the “begging the question” objection fails.
Of course, the argument I have just set forth depends on the assumption of superabundance, and we do not have superabundance in the real world. In the real world, infinite supplies of the requisite substitutes are not available. Accordingly, in the real world, both potential wrongdoers and their victims often have very good reason to suppose that full compensation cannot be paid. This in turn gives real potential wrongdoers a reason to choose (a) over (b), for as a practical matter, producing (b) would be impossible. So if resources are limited, perhaps Hume was right after all—we do need principles of justice to regulate when and to what extent harm could be permissibly inflicted by human agents and to specify what should be done when any of these limits have been breached, regardless of whether we believe in commensurability or incommensurability.
But this need would flow from our contingent inability to compensate, and not from the infliction of the injury itself. For most people, though, it is not the contingent inability to compensate resulting from an absence of superabundance that makes something a moral wrong—it is the infliction of the injury itself. Morality is not merely simply a list of prices for various kinds of conduct, leaving one free to act “correctly” or “incorrectly” as long as one pays the price, whichever one prefers. There is a difference between not committing a moral wrong and committing a moral wrong but fully compensating those injured by it. Those who want to lead a moral life do their best to choose the former, not the latter. While a form of moral thinking is possible under conditions of limited resources and full commensurability, this is not the form of moral thinking in which most of us actually engage. Absent some degree of incommensurability, a need for that kind of moral thinking could not possibly arise. Only in a world where significant and pervasive incommensurabilities were presupposed would the source of our moral concern lie in the infliction of the injury itself and not in our inability to fully compensate for the injury’s effects. Only then would potential wrongdoers have a conceptual rather than a contingent reason not to inflict this injury, to choose (a) over (b). And only then would the role of “morality as we know it” be to guide us toward making what most of us consider to be the correct choice.
What this means is that one important tenet of the moral theories that most of us embrace is that we should avoid wrongfully inflicting incompensable harm on others. What constitutes the “wrongful” infliction of incompensable harm, of course, is to be separately defined, and can vary according to what moral theory one happens to embrace. In other words, inflicting harm is not wrongful merely because the harm is to a carrier of incommensurable value. Whether harm is actively or passively created, self-inflicted, or inflicted entirely by others, it may be morally permissible to inflict it or not. Indeed, the infliction of incompensable harm may sometimes be morally required, as when we punish someone by imposing restrictions on their liberty—restrictions that most people would view as injuring a carrier of incommensurable value. All conduct that results in such an injury is accordingly not necessarily morally wrong. But as long as incommensurabilities are significant and pervasive, all or at least almost all wrongful harms—that is, negative changes in well-being that arise out of moral wrongs—will be to carriers of incommensurable value, at least in part. All or almost all moral wrongs will therefore be to some extent incompensable. And it is this fact that allows us to locate the source of our moral concern in the infliction of this harm itself rather than in what would otherwise be a contingent inability to compensate. 20
The reason why morality as we know it is so concerned with this is that when we wrongfully inflict incompensable harm on others, we deprive them of their moral agency—their ability to use their free will to make choices between carriers of incommensurable value after reflection has allowed them to come to a considered judgment. If we deprive someone of the opportunity to make such a choice, we interfere with their life in a way that cannot be undone, and we accordingly fail to treat them with the respect due every self-aware, feeling, reasoning being. To deprive someone of the ability to exercise their independent moral agency in this way is at least prima facie morally wrong under most people’s conception of morality. If we were to reject incommensurability, in contrast, we could not inflict incompensable harm, and interfering with another’s independent moral agency would not technically be possible, since we could always undo what we have done—at least in theory—given sufficient resources and a way to overcome the epistemological difficulties that often attend determining what, exactly, making full compensation would entail. 21 Only if we accept incommensurability would the commission of the wrong itself, and not our contingent inability to fully compensate those injured by it, be the primary source of our moral concern. Only then could morality as we know it continue to exist.
But does having an incommensurability-based conception of morality mean that in most cases justice cannot be done, for full compensation can rarely be paid? Isn’t naming incommensurability a circumstance of justice a little misleading—isn’t it more accurate to refer to incommensurability as a circumstance of injustice? No, because even though full compensation cannot be paid in many situations, this does not mean that achieving justice is necessarily going to be impossible in such cases. Indeed, it is precisely because full compensation cannot always be paid that we have (at least) two remedies for wrongdoing—compensation and punishment (if full compensation could always be paid, why would we need punishment?). Determining the amount of punishment due is of course sometimes controversial, but most people would agree that one of the factors that goes into this calculation is the extent to which the suffering caused by the wrongdoer remains uncompensated. Only when it is impossible to fully compensate the victim of a wrong and impossible to punish the wrongdoer sufficiently is it true that justice cannot be done, and the number of cases that fall into this category is going to be much smaller than the number in which full compensation is impossible. 22 And even though neither sufficient compensation nor sufficient punishment is going to be available in all cases, and therefore sometimes justice will not be achievable, this does not mean it is inaccurate or misleading to speak of incommensurability as one of the circumstances of justice, for a circumstance of justice is something that makes certain principles of justice necessary, not something that guarantees that these principles and therefore the demands of justice can always be fulfilled.
In any event, what all this means is that incommensurability is not a theory about the moral value of options, acts, states of affairs, or what have you—it is a necessary precondition for moral value as we know it to arise. We are free to believe in incommensurability or not, but if we choose to believe in full commensurability then morality as we know it cannot exist. The choice between commensurability and incommensurability arises prior to the point when the kind of judgments which we associate with morality as we know it can be made. Presupposing incommensurability provides a reason for making such judgments, whereas presupposing full commensurability provides a reason to reject the claims such judgments purport to make. The argument for incommensurability is accordingly the argument for morality itself.
Many theorists, however, believe that even if some carriers of value are incommensurable, all carriers of value are comparable (see, for example, Chang 1997). In other words, they believe that by judging one carrier of value better than another we are not implying that these carriers of value can be measured against a common scale, and therefore we are not implying that we could replace any carrier of value with some measure of another. If this were true, then comparability (unlike commensurability) would not be an extension of superabundance, for it would not imply that some carrier of value could always be substituted for another, even in theory. And under these conditions, morality as we know it could still exist. So to move forward in my argument I need to prove that this combination of views—that values can be incommensurable but comparable—is not possible to maintain. This, then, is the topic that we turn to next.
2. Can values be incommensurable but comparable?
There are two ways in which values might be incommensurable but comparable. First, some values could have lexical priority over others. Only after these prior values were fully satisfied or no longer implicated in any way would other values have any moral relevance in deciding what we should do. These values would be comparable because it would be possible to rank them ordinally, but they would not be commensurable, the argument goes, because they would not have to be measured against a common scale in order to decide which value to promote in a particular case. Second, it is possible that values could be comparable but not commensurable if some carriers of value were infinitely more valuable than others. Exactly what this means is not entirely clear, for it could be understood in several ways. But at the very least, the “infinitely more valuable” relationship, like lexical priority, is intended to suggest that we can ordinally rank some values even when they cannot be measured cardinally according to a common scale. 23 Lexical priority and the “infinitely more valuable” relation are accordingly both designed to accomplish the same thing—to fix a relationship between values without implying that this relationship is contingent on what amount of each carrier of value is available, for as long as an ordering is not contingent on this, there is no need for reference to a common scale, and the two values therefore need not be commensurable. If such fixed relationships were valid, then wherever this kind of relationship existed it would not be possible to replace the carrier of one value in that relationship with the other. Substitutability would be limited, and this would give us all the reason we would need to prefer one state of affairs over another, even if these states of affairs were incommensurable. This, in turn, would go some way toward allowing morality as we know it to exist.
Before I talk about each of these claims in greater depth, however, I want to point out one serious problem that both claims have in common. What I have argued is that for morality as we know it to exist, we must presuppose that significant incommensurabilities are pervasive in our value system. But I have assumed that incommensurable values are also incomparable. If incommensurable values can be comparable, then my argument begins to break down. But to break down enough to allow morality as we know it to exist, lexically prior and infinitely more valuable relationships would have to infect the very core of our value system, and not just populate the fringes. Like the incommensurabilities for which they would substitute, significant examples of these relationships would have to be pervasive. If we could not rank a wide range of values using one or the other of these methods, then even though we would not have the equivalent of full commensurability, we would not have to depart from full commensurability often enough for us to need a comprehensive set of moral principles. On the contrary, we would only need moral principles for dealing with situations where lexically prior values or values that are infinitely more valuable than others collide with values of a lesser kind.
Those who argue for these methods of comparison, however, do not suggest that numerous significant examples of lexically prior or infinitely more valuable options are to be found throughout our value system. Indeed, even Rawls (1999: 38–39) concedes that lexical orderings are only appropriate in certain special situations. Thus, even if it is possible to have incommensurability but comparability given the existence of these methods of comparison, it does not seem that examples of this would be pervasive enough to allow morality as we know it to exist.
But this is not the only problem with these purported methods of comparison. As we shall see, each of these supposed bases for comparability without commensurability has additional problems of its own. 24
2.1. Lexical priority
While no one seems to feel it necessary to explain why lexical priority is supposed to establish comparability without commensurability, the thinking behind this assertion appears to run something like this. If one value has lexical priority over another, additions to the amount of the lesser value do not bring it closer in value to the prior value. On the contrary, the lesser value simply has contingent moral relevance—it has no role to play in our moral reasoning until the prior value is fully satisfied or no longer implicated in any way. Accordingly, when two values are in a lexical relationship, they need not be commensurable, for there is no need to measure the two values against a common scale; one simply applies the priority rule and gets on with it.
Of course, the obvious question here is: Even though we need not measure two competing values against a common scale in order to apply the rule of lexical priority, how do we determine which value is lexically prior to the other in the first place? If we are not doing this by measuring each value against a common scale, then how is this to be done? The method of ordering from which the lexical method draws its name—alphabetizing—cannot help us here, for this tells us nothing about how to rank potentially competing values.
25
And while some theorists claim that establishing lexical priority involves ranking values by “status” rather than by “weight” (see, e.g., Anderson 1997), it is difficult to see how this avoids ranking values according to a common scale (see Williams 2011). Yet if we are determining which value has priority over the other by measuring each against a common scale, then lexical priority does not imply comparability without commensurability. On the contrary, what it implies is this:
Now, the first thing to note about such a relationship is that it is implausible to think that anything like true lexical priority of values actually exists. Do we really think that there are two values such that no amount of one could offset the slightest decrease in the other? Take, for example, perhaps the most famous example of an attempt to rank values lexically, Rawls’s principles of justice as fairness. Under those principles, no amount of social and economic equality could outweigh even the slightest decrease in equality of opportunity, and no amount of equality of opportunity could outweigh even the slightest decrease in basic rights and liberties (see Rawls 1999: 37–39). These claims are not easy to accept, however. We might resist such trade-offs on slippery-slope grounds, but this makes lexical priority just a prophylactic rule, not a true principle of justice. At the very least, most people would say that such trade-offs could be justified in extreme situations, which means that a more accurate graph of values that we treat as having a lexical relationship would show some overlap between the two values at issue, and would therefore look like this:
For our purposes, however, it does not really matter if the first graph or the second is correct. In either case, we have both comparability and commensurability. What we may not have, however, is substitutability. Except in extreme cases where the scales of value overlap, we cannot replace a carrier of the greater value with some amount of the other, and this threatens to break the connection that I have argued exists between incommensurability and morality as we know it. In other words, the existence of such lexical relationships does seem to establish that there are conceptual limits on our ability to compensate for injuries to certain carriers of values. If we cannot compensate certain injuries—if we cannot put a package of values together that can substitute for another—then this would allow morality as we know it to exist.
But this is only true if we view the two values in this relationship in isolation. If both values are being measured according to a common scale, which as I have shown is necessarily the case if we are making these priority decisions by making comparisons, no doubt many other values can be measured against this scale as well. While this is difficult to prove, it is plausible that some combination of values measurable against this common scale could substitute for any amount of value Y. Even if there is only a slight overlap between social and economic equality and equality of opportunity, for example, would that overlap not increase significantly if we were balancing a package containing both greater social and economic equality and a higher level of welfare against equality of opportunity? Would it not increase even more if we could also add equality of resources into the mix? As long as we accept that a common scale exists, this implies that in most cases, at least, values can be packaged in such a way to allow one package to substitute for another. 26 And in this case, the connection between where we locate the source of our moral concern and our acceptance of either incommensurability or commensurability is unbroken, and my claim that we must believe in incommensurability to have morality as we know it remains intact.
Of course, this does not rule out the possibility that we can establish a kind of lexical priority in some other way—that is, without making comparisons, in which case we could have lexical priority that does not imply commensurability and therefore is not inconsistent with morality as we know it. Indeed, those who embrace deontological methods of moral reasoning claim to be doing this all the time. My point is not that we cannot do this—on the contrary, it is essential to my argument that we can do this. What I contend is that we cannot do this and embrace full commensurability at the same time, even if we do employ a deontological method of moral reasoning, for embracing full commensurability would render us unable to morally distinguish between options, values, states of affairs, and so on except on the contingent grounds of non-payment of compensation. I will say much more about this in the final section of this paper, where I will also discuss how we might make priority decisions non-comparatively—that is, without embracing full commensurability—using a deontological method of moral reasoning. But for now, all I want to do is establish that the mere existence of lexical priority rules does not imply comparability without commensurability. To complete the argument that there is no such thing as comparability without commensurability, however, we will need to consider the other way in which values might be thought of as incommensurable but comparable—the claim that some carriers of values can be infinitely more valuable than others—and so that is the issue I shall turn to next.
2.2. Infinite value
The claim that some values are infinitely more valuable than others can be understood in a variety of ways. Suppose the claim is intended to mean that any finite unit of A is worth more than an infinite number of finite units of B. If the units of B have no value, then it does not matter whether there are an infinite number of them or only one—in either case, they would have no value. In this case, the claim reduces to the statement that B has no value, and so all it tells us is that B is not a source of moral concern; it does not tell us how to compare two carriers of value. In contrast, if the units of B each have an equal positive finite value, then the total value of an infinite number of these units would be infinite, so it cannot be the case that one unit of A is more valuable than that—at most, one unit of A and an infinite number of units of B could be equally valuable. 27 And if the units of B have diminishing positive value, then an infinite number of units of B could have either a finite or infinite value. If infinite, then once again, A could not be of greater value; if finite, an infinite number of units of B could be less than the finite value of A, but this would not mean that units of A and B are incommensurable but comparable; it would mean that units of A and B are comparable and commensurable, for the only basis for making such and ordering would be that the finite value of an infinite number of units of B and the finite value of a unit of A could be measured against a common scale. Moreover, even in this case, we could conceivably make up for any shortfall between A and B by adding in some number of units of C or some other value, for there is nothing in the “infinitely more valuable” relation between A and B to rule this out. Finally, if we interpret B to be an infinite number of units of every other value, then the statement becomes totally implausible, for it is difficult to believe that one unit of anything could be more valuable than an infinite number of units of everything else, even if the value of each additional unit was diminishing.
Suppose, however, that we interpret the “infinitely more valuable” relation to mean that any unit of X is infinitely more valuable than any unit of Y, each of which has some positive finite value. In this case, it is unclear what this possibly could mean. Valuable in terms of what? This is a comparative statement, and therefore implies the existence of a reference value. While advocates of comparability without commensurability consistently fail to recognize this, without this reference value, the statement is incoherent. 28 And if there is some reference value underlying such a comparison, then this implies that the value of both X and Y can be measured according to a common scale, which means once again that these two sources of value are not only comparable, they are commensurable too.
There are further problems here, as well. First, do we really believe that some carriers of value are infinitely more valuable than others? This would mean that even though Y had some positive finite value, no matter how much Y we added into the mix, the total value of this would be no closer in value to the value of one unit of X, for that is what being infinitely more valuable means. Can we actually think of something that could be valued in this way? Thinking that one carrier of value is infinitely more valuable than another that nevertheless has some positive finite value seems awfully implausible. What we must really mean when we say that one carrier of value is infinitely more valuable than another is that X is very, very much more valuable than Y, but not really infinitely more valuable, which again implies not comparability without commensurability but comparability and commensurability.
Second, is there any carrier of value that does not exhibit declining marginal value? If not, then it is difficult to see how any single unit of one could be infinitely more valuable than another, for such units do not exist in a vacuum, and must always be evaluated in the context in which they appear. Because they exhibit declining marginal value, each additional unit of X adds less and less to the total value of X. Once X is infinitely more valuable than Y, additional units of X add nothing to the total value of X, because again, this is what being infinitely more valuable means. So we can never say that any unit of X is infinitely more valuable than any unit of Y, because units of X must by definition be differently valued, and some units of X (those marginal units that are added after X has infinite value) must have no actual value, and therefore cannot be worth more much less infinitely more than a unit of Y. The statement that any one unit of X is infinitely more valuable than any one unit of Y simply cannot be true.
Third, the betterness relation is a comparison of the degree of goodness each option has. 29 By definition, any option or carrier of value that is better than another has more good associated with it. Something that is infinitely better than something else in a moral sense is accordingly infinitely good, not merely in relation to the thing to which it is being compared but in an absolute sense as well, because once something becomes infinitely good this is as good as it gets. Even a simple statement like “my children are infinitely more valuable than money” necessarily implies that my children are infinitely good. Most people, however, believe that only God can be infinitely good, and conversely that the only thing that can be infinitely good is God—which suggests again that when people say that X is infinitely better than Y, what they really mean is that X is much, much better, but not infinitely better, for maintaining otherwise would entail embracing certain religious beliefs that would be very controversial indeed.
Finally, even if one carrier of value could be infinitely more valuable than another and therefore not replaceable by any finite number of units of the lesser value, this would not mean that the more valuable carrier could not be replaced by a carrier of some other value. Indeed, since we are presuming that “infinitely more valuable” relations would be pervasive, there would have to be numerous carriers of value that would be infinitely more valuable than others, and therefore infinitely valuable full stop. These alternative carriers of infinite value, in turn, could be substituted for one another, for all carriers of infinite value are by definition of equal value. The fact that X is infinitely more valuable than Y would accordingly not imply that the value of X could not be equaled by some and perhaps many other carriers of value. Despite what the proponents of comparability without commensurability contend, X would still be replaceable, at least in theory. Accordingly, even if pervasive “infinitely more valuable” relationships were possible, this would not produce morality as we know it. The only way to do this is to accept that significant and pervasive incommensurabilities of values do exist.
3. How can we make choices between incommensurable values?
What I have argued is that only if we begin with the presupposition that significant and pervasive incommensurabilities exist can we produce morality as we know it. But there is a problem here. If not all inflictions of incompensable injury are morally wrong, we need a way of deciding which are wrong and which are not. Morality must have something to say about this. If incommensurabilities are significant and pervasive, however, deciding which inflictions of incompensable injuries are wrongs and which are not will often require us to choose between incommensurable options. It seems that incommensurabilities must be significant and pervasive for morality as we know it to exist, but the existence of significant and pervasive incommensurabilities also make it impossible for morality to do the job we need it to do. The same conditions that allow morality to exist also seem to make it powerless to help us decide what is wrong and what is right. Resolving this apparent paradox, or at least suggesting how this apparent paradox might be resolved, seems important if we are to be confident that presupposing incommensurability is indeed necessary for morality as we know it to exist.
One possibility is that when faced with a choice between options of incommensurable value, we simply base our choice on our preferences or desires. But such an option may not be open to us. If, for example, we believe in the desire-based account of value, which provides that something is valuable because we desire it, the fact that two options are incommensurable in terms of value necessarily implies they are incommensurable in terms of preferences or desires as well. In this case, resort to our preferences or desires as criteria for choice will result in the same problem with which we started. If, on the other hand, we believe in the buck-passing account of value, which provides that having value simply means having properties that give us reason to value whatever has these properties, we encounter a different problem. Under this account, the fact that two options are incommensurable in terms of value does not mean that they are incommensurable in terms of desires, but this does us no good unless desires can provide reasons for action. To believe that they can, one would have to believe in the buck-passing account of value but the desire-based account of reasons. Examples of contemporary theorists who embrace this combination of views, however, are difficult to find. Those who accept the buck-passing account of value all seem to embrace a value-based account of reasons, under which desires do not provide reasons for action except perhaps in some trivial or exceptional cases, and they reject the idea that a functional set of reasons for action could be provided by desires alone. 30 Under the value-based account of reasons, however, we cannot have reasons for engaging in one action rather than another if we cannot compare the value of the states of affairs those actions would produce. 31 So if one cannot believe in the buck-passing account of value and the desire-based account of reasons at the same time, then it is difficult to see how one could rationally choose between two incommensurable options no matter what account of value one happened to embrace. 32 In other words, as Raz (1986: 345) says, it seems that “reason cannot advise us how to choose between options which are incommensurate.”
Of course, there are ways in which we could make choices between options without resorting to reason. We could simply choose randomly, or on the basis of habit, or convention, or faith. But the availability of these methods of choosing does not solve the paradox that an incommensurability-based conception of morality seems to create. These methods of choosing merely establish that we can make decisions in certain situations without help from morality. What we need to explain, however, is why the conditions that make morality possible do not also render it impotent in the very situations we need it most. Without that explanation, the paradox that arises from my argument in favor of significant and pervasive incommensurabilities necessarily remains.
The first thing to note in relieving some of the pressure of this paradox is that it is not true that morality is totally useless when it comes to making choices between incommensurate options. Morality can still tell us what counts as a reason in favor of each option and how important in a non-relative sense that reason seems to be, even if the reasons in favor of each option cannot be compared. It therefore allows us to distinguish valuable options from options that have nothing to recommend them. Indeed, this is what allows Raz to argue that while reason cannot assist us in choosing between incommensurable options, this does not mean that whatever action we ultimately select is irrational. On the contrary, he says, when two options are incommensurable, this means there are reasons supporting each choice. Moreover, by definition, both sets of reasons remain undefeated, so it is not possible to say the reasons supporting one choice are better than, worse than, or equal to the reasons supporting the other. As a result, either choice is rational, for each can be based on an undefeated reason, and this is all that rationality requires (Raz 1986: 338–339).
This, however, would seem to make choosing between incommensurable options no different from choosing between options of equal value, for in each case we would have insufficient reason to make the choice. Yet as Raz himself recognizes (1986: 331–332), a choice between incommensurable options and a choice between options of equal value are very different kinds of choices indeed (for a similar observation, see Broome 1991: 7). In the latter case, we are indifferent between the options. But in the former case, we are not indifferent—in many cases, despite the fact that the options are incommensurable, we feel that the choice between them matters very much. It is not enough to say that if there are undefeated reasons on both sides of the equation, then morality permits us to make either choice. What we want is not moral permission—we want morality to give us a recommendation even in these circumstances. It is how morality might generate this recommendation that seems so mysterious.
Raz does offer one observation that might provide some insight into how such a choice might be made: He notes that when faced with a choice between maintaining the status quo and doing something else when these two options are incommensurable, people will typically choose to maintain the status quo (Raz 1986: 346). It is not clear whether Raz is merely noting the effect of the common status quo bias here or actually endorsing this as a rational principle of decision-making, but even if he is doing the latter, it is hard to see how this could be interpreted as a substantive rule of morality. At best, it is an evidentiary rule, a rule regarding the burden of proof, which says that unless there is a positive moral reason to do something else one should stick with the status quo. What we are looking for, however, is not a mere evidentiary rule, but a way of giving morality a substantive role in guiding us toward what we should do. Besides, in many situations, sticking with the status quo will not be an option. One will have to choose between two affirmative options, and Raz’s rule of evidence cannot help us do that. In any event, unless we can show how morality can help us here, we have still not explained how the conditions that render morality possible do not also rob it of much of its utility as a guide for human conduct, its reason to exist.
But there is a way to solve this paradox by drawing on Raz’s work, although Raz does not seem to recognize that this is what he has done. In arguing for incommensurability, Raz argues for what he calls constitutive incommensurability. He gives an example of an offer to a set of parents to trade their child for a sum of money. Raz argues that most parents would regard such an offer as abhorrent, for the very idea of comparing the value of their child to a sum of money, no matter how large, is inconsistent with the concept of parenthood, and they could not make such a comparison and still conceive of their relationship with their child in the same way. The explanation for this cannot simply be that children are worth more than any amount of money, since then people would be willing to buy children but not sell them, and people typically react with the same abhorrence to both kinds of offers. The only way to explain this reaction is that for most people, the two options are both incommensurable and incomparable—one is neither better than, worse than, nor equal to the other. For people who understand and are committed to what it means to be a parent, such comparisons simply cannot be made.
What Raz does not seem to notice is that this example not only proves that constitutive incommensurabilities exist; it also shows that morality must be able to give us some guidance on how we should choose when confronted with incommensurable options. It is clear that most people who might be approached as prospective buyers or sellers of children feel that the morally correct choice in this situation is to reject all such offers, and this cannot be simply the result of a status quo bias, for characterizing these decisions as the mere application of an evidentiary rule simply does not capture the essence of what is going on. These people do not reject the idea of trading money for children because they have no positive moral reason to engage in such a trade, but because they believe that such a trade would be morally wrong. The moral theory they embrace demands that they reject such offers. But if the options offered by such a choice are both incommensurable and incomparable, how can this be? How can morality tell us anything in this situation about what choice we should make?
The answer, I think, is that the pair-wise comparison is not the only available method of moral reasoning. Rather than looking for a betterness relation between the available options and then selecting whichever option is better, we can instead develop a set of necessary and sufficient criteria for judging what course of conduct morality requires and then apply those criteria to the given situation (see Rawls 1999: section 21, especially 106–108). If both options meet these criteria, then the criteria may need to be further refined. Alternatively, as long as these choices are not pervasive, satisfying rather than maximizing criteria may be sufficient, and we can simply regard both options as morally permissible. In any event, as long as sufficiently detailed criteria can be developed, and we are careful to ensure that these criteria are not themselves derived by making comparative moral judgments, then morality can help us to decide which option to select even though it is not possible to make a value comparison between the states of affairs they will produce.
Let us return then for a moment to our judgment that we ought not to exchange our children for some amount of money—how might this judgment be reached without engaging in comparisons of any sort? Well, we might believe that the respect to which the special relationship between children and their parents is entitled demands that we not treat that relationship as amenable to purchase through a means of exchange. To contemplate selling or buying children would therefore violate this duty regardless of what other good might flow from engaging in such an exchange, including good to the children themselves. No comparison is possible between the value of the amount of money on offer and the child at issue because we could not contemplate such an exchange and still think of the value of the parent–child relationship in the same way. But no comparison is necessary either, for no judgment of betterness is required here—our moral beliefs here are derived without having to make teleological comparisons and function instead as a pure Nozickean side-constraint—they demand that we reject the very idea of such an exchange. Rightness instead of betterness provides the criteria for moral choice, and under these criteria children are not to be exchanged for money in any amount.
I recognize that the above statement needs some qualification. Obviously, people make choices all the time that appear to involve a trade-off in some sense between children and money. I take a job in another city that pays much more than the job I currently have, even though this means I will be able to see my children only on weekends, and yet this choice does not necessarily seem to require redefining constitutive elements of my relationship with my children. Indeed, Chang (2001) makes this very point in the course of criticizing Raz’s view. She thinks this means we must be making comparisons here, and if we are making comparisons here, we must be making comparisons even when presented with an offer to sell or buy children. There is another explanation, however, which she does not consider. While I have in some sense exchanged time with my children for money in the “better job” example, I have not in any way attempted to transfer my parent–child relationship to another, which is the choice that the “selling/buying” example posed. The difference may be subtle, but it is fundamental. Merely because we are making comparisons in one example does not mean we are making comparisons in the other. So the better job example does not negate the possibility that we are making our choice in the buying/selling example by using some set of non-comparative criteria.
What, then, do we say about paying ransom to kidnappers or paying fees in connection with adopting children? In some sense, at least, these cases do seem like instances where we are exchanging money for a parent–child relationship. But I am not sure this is correct. In the kidnap and ransom case, it is not the parents who are monetizing the parent–child relationship, it is the kidnapper who is doing that. The parents are not putting an exchange price on their child as long as they do not attempt to bargain over the amount of ransom to be paid once that amount is at all within their means. Think of how their child would react if her parents did try to negotiate the amount of ransom down once it was within their means—she could never think of her relationship with her parents in the same way. 33 And no one would insist that parents offer more than the kidnapper demands, up to their ability to pay, just to be sure they do not suggest their relationship with their child is worth any less. Money just has no business being a relevant factor here on the parents’ side, at least if one wants to maintain a parent–child relationship of a certain kind.
With regard to the adoption case, in turn, there we are simply reimbursing costs incurred by others rather than monetizing the parent–child relationship. I recognize that this may seem like a fine line to draw, but there does seem to be a line here nevertheless. This is why there is something troubling about a set of putative adoptive parents who choose between adoption agencies because one is cheaper than the other even though both are within their means. Once again, it seems like cost has no business being a relevant factor here, at least unless all other potential choice criteria are equal, which seems unlikely, and perhaps even if all other choice criteria are equal. Which agency one chooses, after all, determines which child one will ultimately receive. Who would want to tell their adopted child that they have him or her and not some other child because they went with the cheaper agency, even when all other factors were equal? Even parents who deliberately choose the more expensive agency, at least if they do so merely because it is more expensive rather than taking expense as some sort of signal of hidden quality, are doing something unseemly, for choosing children because they are expensive seems as wrong as choosing them because they are cheap. So it is not merely that the parent–child relationship is worth more than any amount of money, it is that the parent–child relationship and money just do not mix. The values that money and the parent–child relationship carry are, as Raz (1986: 345–54) says, constitutively incommensurable. Morality nevertheless provides guidance on what to do here, and what it tells us to do is to keep monetary considerations out of the decision-making process altogether.
Of course, a moral side-constraint that says do not monetize the parent–child relationship seems easy to derive non-comparatively. When value A would change in nature if brought together with value B, these two values should be kept apart, a constraint that reflects what we might refer to as a kind of Heisenberg principle of morality. But coming up with a principled set of moral criteria that gives us guidance on what to do in a wide range of decision situations is no easy task, and there is certainly no agreement yet on how we should go about deriving such criteria. But we do seem to be able to come up with such criteria as we go intuitively. If we continue to believe these intuitions are correct even after we subject them to considered reflection, then articulating the underlying moral theory that is generating these criteria must be possible. 34
A useful analogy might be drawn here between the judgments “morally right” and “funny.” We may decide that one thing is funnier than another by comparison, but we cannot use comparisons to determine whether something is funny full stop. In other words, something is either funny or it is not—a certain non-comparative threshold must be reached before we can begin to classify something as funny; up until that point, at least, comparisons are neither necessary nor appropriate. This point, in turn, is determined by applying certain necessary and sufficient criteria of funniness. 35 While it may be difficult to articulate all of these criteria in advance, and may even be difficult to articulate them in retrospect, we do seem to be able to recognize them intuitively. Indeed, not only can we recognize them intuitively, but we can also recognize when they have changed, which is why, for example, racist jokes that once struck many people as funny do not often strike them as funny anymore. In any event, the difficulty in setting forth the principles from which these criteria are derived and modified has not convinced us that such principles do not exist. And if a set of principles for determining what is funny does exist, a set of principles for determining what is morally right even when comparisons are impossible is likely to exist as well.
I should note, however, that in response to the above argument, a few people have argued that it is self-defeating in the following sense: If criteria for rightness can tell us how to choose between incommensurable options, they can also tell us how to choose between options that are equal in moral value, and this is the conceptual position in which the deontologist who embraces full commensurability finds himself. In other words, what my own argument establishes is that a form of the “begging the question” objection I dismissed earlier is essentially correct. Even if the choice between (a) not committing a wrong and (b) committing the wrong but fully compensating everyone injured by it cannot be made by using criteria for betterness, it can be made by using criteria for rightness, and this can be done without comparing the states of affairs the various options in question would produce. There is therefore no paradox of incommensurability and the problem that I claim full commensurability raises for morality as we know it disappears.
But once again, I do not think this is correct. As I have shown, in a world of full commensurability, the states of affairs produced by options (a) and (b) would in all material respects be morally indistinguishable, even considering how each state of affairs was produced. And if these two states of affairs are morally indistinguishable, we would necessarily be morally indifferent between options (a) and (b). It is therefore hard to see how we could at the same time believe that choosing one option over the other is what morality requires, no matter what criteria we use or how those criteria were developed. Indeed, once we decide that two options produce states of affairs that are morally equivalent, it is tautologically true that there can be no work for morality left to do even if there are consequence-independent criteria for rightness that allow us to develop moral rules non-comparatively. And morally equivalent states of affairs would (as a conceptual matter at least) be endemic in a world where we believed in full commensurability. Only in a world where we believed in incommensurability could the search for deontological criteria for rightness make any sense, because only then would there be a decision to be made between (a) and (b)—between not committing a wrong and committing that wrong but fully compensating those injured by it, conceptually if not actually (although in most cases actually too), and so only then would the question of what is morally right to do remain open to be decided by non-comparative means, and only then would we be able to put non-comparatively derived moral criteria to any use.
I want to make clear, however, that accepting incommensurability does not mean we must abandon consequentialism and the balancing of moral value that it entails as a method of moral reasoning altogether. While arguments for incommensurability are often thought of as indictments of utilitarianism and other forms of consequentialist moral reasoning, this is not how my claims should be understood. On the contrary, my view is that people are free to accept incommensurability or not, so my argument poses no problem whatsoever for consequentialists who have chosen to believe in full commensurability as long as they are willing to accept the consequences of embracing such a belief. Moreover, even those who are unwilling to make this concession are not necessarily forced to reject consequentialism altogether. While consequentialist methods of moral reasoning rely heavily on comparisons regardless of whether we are comparing the moral value of acts, or of the states of affairs they produce, or both, there is no reason why we must abandon consequentialism just because the required comparisons cannot always be made, for consequentialist moral reasoning may be able to exist inside a larger over-arching deontological structure that supplies the rule of decision-making when the normal tools of consequentialist reasoning are unavailable. I will not be able to say more about how such a hybrid form of consequentialism might work, for this is too big a topic to explore here, but I do want to make clear that nothing I have said in this paper would foreclose the development of such a theory.
I also want to make clear that while deontological methods of moral reasoning offer the best hope for deriving and identifying morally relevant criteria for choosing between incommensurable options, we must also recognize that the fact that our method is deontological does not necessarily mean we can easily avoid the need to make comparisons between incommensurable values. As I have argued at length elsewhere, one can—and perhaps sometimes must—engage in balancing considerations against one another even in the course of developing deontological rules (see Reiff 2009, especially 27–31). Once these rules have been developed, they can arguably be applied without balancing other considerations against them. But the scope of each rule is always subject to interpretation, and this means that those who engage in deontological moral reasoning will often be faced with definitional disputes regarding the scope of protection a particular right entails. Does the right to life, for example, preclude abortion? Does the right of a woman to control her own body demand that abortion be permitted? Indeed, if these definitional disputes are not to be decided by balancing, it is difficult to see how they are to be resolved. Consequentialists are accordingly not the only ones who must explain how decisions can be made when balancing is not possible.
While I do not think this problem is insurmountable, it is nevertheless too complex and controversial an issue for me to adequately address here. For now, all I can say is that we seem to be able to recognize the requisite necessary and sufficient criteria in a wide range of decision situations using our moral intuitions. So the paradox that an incommensurability-based conception of morality appears to create is not actually a paradox at all—our moral practice demonstrates that we in fact use the very incommensurability-based conception of morality I argue for in the first two parts of this paper to guide us in difficult choice situations all the time. What we have not yet done is articulate fully how we do this—but such an explanation must exist, for we do indeed do this. Articulating these non-comparative necessary and sufficient criteria for moral choice is simply what any theory that hopes to bring our moral intuitions into the light and explain how morality as we know it operates must set as its central task.
Footnotes
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Acknowledgements
This paper is dedicated to Hillel Steiner, who inspired me to develop and defend my views on incommensurability. Earlier versions of this paper were presented at the Safra Center for Ethics at Harvard University, at the Tenth Conference of the International Society for Utilitarian Studies at the University of California, Berkeley, and at the University of Manchester Center for Political Theory. I want to extend my thanks to all those in attendance for their comments and suggestions. My special thanks go to James P. Crutchfield and Mary Leng for helping me with some concepts in the mathematics of infinity, and to Arthur Applbaum, Harriet Davidson, Matthew Kramer, William Lucy, Hillel Steiner, Peter Vallentyne, and two anonymous referees for this journal, each of whom provided me with numerous thought-provoking and helpful comments and suggestions after reviewing earlier versions of the paper.