Moral Rules Russ Shafer-Landau Ethics, Vol. 107, No. 4. (Jul., 1997), pp. 584-611. Stable URL: http://links.jstor.org/sici?sici=0014-1704%28199707%29107%3A4%3C584%3AMR%3E2.0.CO%3B2-B Ethics is currently published by The University of Chicago Press.
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/ucpress.html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers, and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take advantage of advances in technology. For more information regarding JSTOR, please contact
[email protected].
http://www.jstor.org Tue Nov 13 16:08:42 2007
Moral Rules*
Russ Shafr-Landau The traditional conception of ethical theory sees it as the project of developing a coherent set of rules from which one can infer all determinate moral verdicts. I am not optimistic about the prospects for constructing such a theory. To explain this pessimism, we need to understand what moral rules are and what roles they might play in ethical theory. I. THE NATURE OF MORAL RULES
A moral rule states or expresses a relation claimed to obtain between a moral property and other, grounding properties that are correlated with its instantiation. The correlation between moral properties and their grounds is always alleged to be universal. In other words, moral rules either obviously do or can be easily translated to take the following form: for every x, if Gx, then Mx, where 'G' stands for some grounding property or properties, and 'M' stands for some moral property. The relevant domain is everything possibly subject to ethical assessment. It should be clear from this schematic representation that the universality of a moral rule is independent of its scope. The scope of a moral rule is determined by the concreteness of its incorporated grounding properties. The moral rule that prohibits all killing has a broader scope than one that prohibits all killing except that done in self-defense. So universal rules may be very narrowly drawn, or quite highly general.1 It should also be clear that no metaethical commitments follow from the claim that moral rules are universal. Moral
* My thanks to Jack Bricke, Ann Cudd, David Gill, Kinch Hoekstra, Don Marquis, and Henry Richardson for identifying weaknesses in earlier drafts. Work on this article was supported by a General Research Fund grant (3756-20-0038) from the University of Kansas. 1 . See R. M. Hare ("Principles,"Proceedings ofthe Aristotelian Society 7 3 [1972]:1 - 18) and Martha Nussbaum (Love's Knowledge [Oxford: Oxford University Press, 19901, pp. 38-39, 67-68) for instructive remarks on the distinction between generality and universality. Ethics 107 (July 1997): 584-611 0 1997 by The University of Chicago. AU rights reserved. 00141704/97/0704-0004$02.00
Shafer-Landau
Moral Rules
585
rules are universally quantified conditionals. But their logical form implies nothing about whether such conditionals are to be construed realistically, noncognitively, or somewhere in between.* Likewise, the logical form of moral rules by itself is silent on the question of their stringency. The stringency of a moral rule refers to its weight, relative to other considerations, in determining the moral status of a situation. In this regard, a moral rule can be either absolute or prima facie. If absolute, it can never be overridden, and it overrides all possibly competing considerations; there are no circumstances where its force is abridged. Prima facie rules, on the other hand, create rebuttable presumptions that something possessing the grounding property specified by the moral rule will also possess the linked moral property. The prima facie nature of a moral rule does not undermine its universality. Rather, we should read such rules as follows: for every x, if Gx, then presumptively M X . ~This is surely true to the spirit of Ross's work, since he thought that, for example, every promise carried with it some to-be-doneness; every act of maleficence generated some presumption against its performance. Thus for every case, the instantiation of the relevant grounding property in a prima facie rule makes some contribution to the instantiation of the linked moral property. This is opposed to a view of prima facie rules that sees 'prima facie' as a probabilistic sentential operator. On this reading, a prima facie moral rule tells us only that the instantiation of its grounding property makes it (to some degree) probable that the correlative moral property is instantiated. But this fails to capture the idea that the instantiation of a grounding property, if it is covered by a moral rule, always makes some negative or positive contribution to the situation's moral status. This contribution may be slight, depending on the circumstances of a given case. But it does not reduce to nothing; if it did, then the grounding property does not properly find itself in a moral rule. The probabilistic reading is compatible with the contribution being zero, and so we need to look elsewhere for a proper understanding of what it is for the instantiation of a grounding property to create a moral presumption. 2. Thus my talk of "moral properties" is intended to be metaphysically neutral. Realists will take such talk literally. For an antirealist, a moral rule is genuine just in case the instantiation of a nonmoral grounding property invariablyjustifies assent to a particular moral evaluation. Those with antirealist preferences are invited to make the necessary substitution in all that follows. 3. See Donald Davidson ("How Is Weakness of the Will Possible?"in his Essays on Actions and Events [Oxford: Oxford University Press, 19801, pp. 21-42) for a criticism of this way of putting the logical form of a prima facie rule. Despite this disagreement, however, we agree on the key issue, namely, that the grounding property in any wellformed prima facie rule must be universally morally relevant.
586
Ethics
July 1997
Another view of prima facie rules sees the presumption that they create as an epistemic one: the instantiation of a grounding property invariably makes it such that, absent defeaters, one is epistemically justified in believing that the correlative moral property is instantiated. An epistemic reading is problematic, however, because the grounding properties that serve as invariably good epistemic clues might not invariably contribute to the instantiation of the linked moral property. For instance, it may be that whenever one sees someone giving money to a panhandler, then that is a good (albeit defeasible) epistemic reason for thinking the person generous. However, it may be that giving such money is not, as a matter of fact, always a good-making (better: a generosity-making) feature of actions. If there is just a single grounding property whose instantiation is invariably "evidential," but only variably "contributory," then the epistemic reading cannot be a fully satisfactory account of the nature of a prima facie rule. A prima facie rule is genuine (true, legitimate,justifiable, correct) if and only if its grounding property is invariably morally relevant. On my view, a grounding property is invariably morally relevant if and only if its instantiation always makes the same kind of concrete contribution toward the instantiation of some moral property. For the grounding properties of prima facie rules, any such contribution may be overridden by competing considerations. For instance, if promise keeping is a prima facie duty, then in every case, it contributes something valuable to an action and makes it obligatory unless some weightier conflicting consideration is present. In the presence of weightier competing reasons, the positive value of fidelity is not extinguished, just overridden. A property's failure to impart the same kind (not the same degree) of influence in every case in which it is instantiated bars it from serving as a grounding property in a moral rule. The classification and discussion of moral theories I offer below depend on viewing prima facie rules in just this way. 11. ROLES FOR MORAL RULES Moral rules can play at least two roles. The first is investigative-our awareness of rules might assist us in coming to appreciate the ethical status of the options we confront. I think that some rules, suitably understood, can perform this function. But strict adherence to rules may also hinder a sensitive appraisal of the ethical options before us. Literature and life provide obvious examples of those whose rigorism quells moral imagination and precludes compassion when it is genuinely needed. My aim here is not to focus on the epistemic or investigative role that rules might play, but instead to concentrate on their role as elements in fixing the truth (endorsability, for antirealists) of some moral conclusion. A standard model of this role-the subsumptive,
Shafer-Landau
Moral Rules
587
or absolutist model-requires that conclusory ethical assessments be deducible from a moral rule conjoined with a description of a concrete case. The rule cites a grounding property that is correlated with some moral feature, and the case description certifies that the grounding property is instantiated by the act or trait under scrutiny. Ross's familiar model of prima facie rules offers a very different picture.* For Ross, particular judgments cannot be deduced from a conjunction of rule and case description. Rules tell us what kinds of actions can be morally obligatory, but the deontic status of an action is not entailed by its instantiating the grounding property of a prima facie rule. Though the grounding properties in prima facie rules invariably create certain moral presumptions, these may be overridden, depending on the constellation of other grounding properties instantiated in a given case. The absolutist and the prima facie models each offer a very different conception of what the appropriate role for moral rules might be in ethical assessment. The prima facie model limits this role to that of identzjjing morally relevantfeatures. According to this model, certain features of a situation are morally relevant in virtue of being subsumable under a moral rule, that is, in virtue of instantiating the grounding properties specified by a prima facie rule. The absolutist model accepts the idea that moral rules can identify morally relevant features. It claims more for itself, however, by endorsing the idea that moral rules can (in conjunction with specific descriptions)Jix a moral verdict in particular cases. One way to distinguish among the many possible versions of each model is on the basis of coverage. One prima facie theory covers more than another just in case it identifies a greater number of grounding properties. At the limit, a prima facie theory may claim to cover every morally laden case. This is another way of saying that a feature of the world is morally relevant if and only if it instantiates a grounding property in a prima facie moral rule. Such a theory sees the set of prima facie moral rules as delimiting the set of morally relevantfeatures. Weaker theories allow for the possibility of grounding properties that cannot be captured by prima facie rules. These are properties that sometimes, but not always, create some particular moral presumption. If there are any such, then the set of prima facie rules cannot delimit all morally relevant features. Absolutist theories differ in their own way on matters of coverage. One absolutist theory covers more than another just in case it fixes a greater number of ethical verdicts. The strongest version of absolutism would claim that every determinate ethical verdict can be deduced
4. W . D. Ross, The Right and the Good (Oxford: Clarendon, 1930).
588
Ethics
July 1997
from an absolute rule. Weaker theories would allow for a range of conclusory ethical judgments that are not deducible from absolute rules. It is important to note that a commitment to absolutism does not entail a commitment to thoroughgoing ethical determinacy. The absolutist model purports to tell us how moral verdicts are fixed when they can be fixed; it does not by itself commit us to a view of whether such verdicts are always a~ailable.~ Indeed, advocates of any but the strongest version of absolutism might claim that those situations not covered by an absolute rule are precisely those that are morally indeterminate. Both the absolutist and prima facie models have their special attractions and difficulties. Absolutism promises justificatory security once the relevant moral rules are identified. Nothing convinces like a good moral deduction. But complex or novel moral cases invariably arise where we feel the need to alter absolutist rules, thus undermining confidence that we have ever really accurately captured their content. It seems easier to identify prima facie rules, largely because they aren't expected to stand as major premises in deductive syllogisms. But this may be a Pyrrhic victory, because the prima facie model appears to leave us no means of justifying our intuitivejudgments about particular cases. A moral rule that overrides its competitors in one instance may be overridden in another; citation of a prima facie rule is rarely determinative of the moral verdict to be reached in the case at hand. 111. IDENTIFYING MORALLY RELEVANT FEATURES Particularists are those who deny that moral rules can fulfill the three roles just enumerated. The strongest form of particularism would deny that rules can even identify morally relevant features. A weaker form would allow rules to identify morally relevant features, but would deny that rules alone can identify every morally relevant feature. The weakest form of particularism would allow rules a delimiting role, but deny that rules could fix the moral status of any situation. If the strongest version of particularism were true, then both the prima facie and the absolutist models would have to be abandoned. For both models assume that the grounding properties in their rules invariably identify features whose (presumptive or absolute) moral relevance is fixed. Recently, Jonathan Dancy has rejected both models of rules by endorsing this strongest version of particularism: "We cannot in general say, where an action has moral property M in virtue 5. Though sympathetic with many of their claims, here I disagree with Mark Johnson (Moral Imagination [Chicago: University of Chicago Press, 19931, chap. 7) and Albert Jonsen and Stephen Toulmin (The Abuse of Casuistry [Berkeley: University of California Press, 19881, chaps. 1, 2), who seem to assume that practitioners of rulebased ethics must view their enterprise as one of parceling out a thoroughly determinate moral order.
Shafer-Landau
Moral Rules
589
of having nonmoral properties A, B and C, that the presence of A, B and C count generally as reasons for calling actions M . . . . It is not the case that where my reasons for calling an action good are that it has properties ABC I am committed to calling any other action which has the properties ABC good."6 And here is David McNaughton: "Whether or not one particular property is morally relevant, and in what way, may depend on the precise nature of the other properties of the action. . . . We cannot know, in advance, what contributions any particular nonmoral property will make to the moral nature of an action."' In short, there are no moral rules because there are no features of the world that are always morally relevant, much less always morally relevant in the same way. Yet rules purport to identify grounding properties that are universally and uniformly morally relevant, that is, always such as to give rise (conclusively or presumptively) to a particular moral property. Consider a candidate prima facie rule: one ought not to cause pain to others. As a prima facie rule, this purports to specify a consideration that has at least some moral weight in all situations where pain is imposed. But, says the particularist, we can imagine some scenarios in which the injunction entirely lacks weight: where others are being justly punished, or where an experiment on pain is being conducted with the informed consent of the participants. The particularist argues that for any imaginable prima facie rule, we can envision scenarios in which the grounding property expressed in the rule does not yield even the presumptive moral weight claimed for it.* If this is so, then prima facie rules cannot exist. And if the grounding properties in such rules can't have even presumptive weight, they can hardly have conclusive, overriding weight. So the failure of the prima facie model would signal the failure of the subsumptive, absolutist model as well. What can be said on behalf of a particularism that denies the existence of universally relevant grounding properties? Usually the case begins by noting the extreme complexity of our ethical life, the possibilities of surprise, novelty, and uniqueness that stand as a barrier to wholesale ethical codifiability. It often proceeds by trying to show that there are exceptions to many putatively universal injunctions. Dancy argues in just this way. He can readily draw cases where it is wrong to return what's borrowed, right to intentionally inflict pain, wrong to please another, right to lie to a n ~ t h e r . ~ 6. Jonathan Dancy, "On Moral Properties,"Mind 90 (1981): 367-85, esp. pp. 376-77. See also Dancy, "Ethical Particularism and Morally Relevant Properties,"Mind 92 (1983): 530-47, esp. pp. 533-34,541-42, and Moral Reasons (Oxford: Basil Blackwell, 1993), pp. 92-107. 7. David McNaughton, Moral Vkion (Oxford: Basil Blackwell, 1988), p. 193. 8. See Dancy, Moral Reasons, pp. 60-66, for this strategy. 9. Ibid., pp. 60-61.
590
Ethics
July1997
Dancy's aim is to show that these grounding properties are only variably relevant. The problem is that examples alone cannot establish this. Consider Kant's inquiring murderer. If we assume that we may lie to him, we are faced with two choices. The first has it that there is nothing even presumptively right here about telling the truth; though crucially important elsewhere, truth telling can entirely alter in its moral value depending upon context. The second alternative claims that truth telling is always positively valuable, though here its importance is overridden. Strong particularists assume that for any putative grounding property, there will be some scenarios for which the first alternative is true. The difficulty with such a view is that we lack an account of how to decide in a given case whether a property's omnipresent relevance is being overridden, or whether the property isn't always relevant. It seems to many-certainly to Ross-that the prima facie value of truth telling, beneficence, and fidelity is selfevident. One doesn't undermine such confidence merely by drawing cases where it is appropriate to lie or scrimp or be faithless. Almost everyone will agree that such behavior is sometimes called for. But this consensus underdetermines a choice between seeing the value of the particular properties as overridden, or as cases in which the value these properties usually possess is absent. Dancy's examples may tempt one to the view that sees a property's value as entirely reversible depending on context only because he has focused on candidate properties that are very generally described. But once we narrow the scope of the grounding properties, it is not very difficult to identify some that are universally and uniformly relevant. Taking pleasure at another's pain, or another's undeserved pain, or from the performance of a wrong action is always presumptively wrong. Likewise, it is always prima facie wrong to be cruel; one's cruelty always contributes negatively to a moral assessment of one's actions. Killing a healthy and happy child is always prima facie wrong. Absent justification, every such action is wrong, and this is just what it is to be prima facie forbidden. There may be very rare circumstances where such action is, all-things-considered,justified, but its permissibility must be a matter of overriding the wrong-making features present in each such action. There isn't any reason to restrict ourselves to the deontic realm, since our ethical concerns are not limited to determining whether actions are forbidden, permissible, or obligatory. Giving to another from unselfish motives, especially if one has very little oneself, is always relevant to determining whether one acts generously. Acting intentionally to cause offense is always relevant, and in the same way, to determining whether an action is rude. If a person saves the lives of others while putting herself at grave risk, all the while knowing the dangers involved, then her action is prima facie courageous. Dancy and Mc-
Shafer-Landau
Moral Rules
59 1
Naughton deny the existence of any moral rules because they deny the existence of their essential building blocks-grounding properties of universal and uniform moral relevance. The plausibility of any one of these examples is sufficient to undermine this strongest sort of particularism.lo IV. DELIMITING MORALLY RELEVANT FEATURES The Case against Delimiting If these or other examples succeed, then prima facie moral rules exist. The weakest of the roles for moral rules can be fulfilled-such rules can identify uniformly relevant grounding properties. But much more needs to be done before knowing whether rules are capable of delimiting the set of morally relevant features. The delimiting thesis can be put as follows: a feature of the world can acquire moral relevance if and only if it is covered by a moral rule, that is, if and only if it instantiates a grounding property in a moral rule. A property can be a grounding property in a moral rule if and only if it is invariably relevant. So the delimiting thesis is true if and only if every morally salient feature of the world instantiates an invariably morally relevant property. Properties are invariably morally relevant if and only if every one of their instantiations is morally relevant. Falsifying the delimiting thesis appears to be simplicity itself. One need cite just a single example of a variably morally relevant property,
10. Some may balk at my use of thick concepts to construct counterexamples to the strong particularist thesis. Trepidation may arise from two sources. First, thick concepts have a relatively small set of grounding properties that are definitionally associated with them, and so differ from more general terms of moral appraisal ('right', 'wrong', 'good', 'bad', 'praiseworthy', etc.) that figure in most "fundamental" moral rules. And second, some of the grounding properties that I cite (cruelty, unselfishness) are themselves moral properties, and so cannot be grounding properties in the strict sense. To take these in reverse order: some examples I provide do not employ moral properties as grounding properties, and so escape whatever force this second objection may have. In any event, this second objection misses the mark, because rule-based grounding properties in the strict sense are simply those whose instantiation is invariably relevant to the instantiation of some other property. There is no reason to suppose that moral properties cannot be invariably relevant in this way. If, for instance, cruelty is in every instance a wrong-making feature, then it makes perfect sense to express this fact in a rule (very likely a prima facie one). That the rule relates two moral properties to each other does not make it any less a moral rule. Nor does it matter that many of the examples I supplied rely on grounding properties that have some definitional association with their correlated moral properties. Moral rules exist just in case the instantiation of a moral property is invariably related to that of some other property (or properties). Thick properties seem to me especially good illustrations of this possibility. Though it is admittedly easier to construct rules from thick rather than thin moral properties, the crucial question is whether we can construct them at all; that we can is sufficient to undermine strong particularism.
592
Ethics
July 1997
a property whose value alters with context. Consider what seems an obvious case: generosity. It is ordinarily admirable to be generous, but there seem to be exceptions-spoiling a child, overindulging an exploitative relative, aiding an assassin. The moral salience of generosity certainly appears to be dependent on context, to be variably relevant. If the appearances don't deceive us here, then there is no prima facie duty to be generous. It would sometimes, but not always, be presumptively right to be generous. Thus some actions would be morally valuable independently of being derivable from a rule. The delimiting thesis would be false.
First Reply: Intrinsic Value and Moral Relevance The first problem with offering such examples is the now-familiar one of knowing whether an example refers to a property possessed of variable relevance, or to a property whose invariant value is being overridden. Perhaps generosity even to layabouts or would-be murderers has some good to it, even if overridden by the ill effects of such kindness. How could we know? There are three options. Generosity is uniformly morally valuable only if it invariably yields good consequences, is invariably a constitutive element in some larger valuable action or plan, or is intrinsically valuable. Generosity does not invariably lead to good consequences and does not always figure in a larger valuable activity. So it is invariably valuable only if it is intrinsically valuable, that is, valuable for its own sake, independently of context or consequences. Generosity is a typical example. Few if any properties are such that their every instantiation is consequentially or constitutively valuable. So every, or almost every, invariably valuable property must be intrinsically valuable. If the delimiting thesis is true, then every valuable property is invariably valuable. So if the delimiting thesis is true, every, or almost every, valuable property is intrinsically valuable. A property is intrinsically valuable if and only if every one of its instantiations is intrinsically valuable. One might think the delimiting thesis false if one could show that there are many extrinsically valuable properties (properties whose instantiations were either consequentially or constitutively valuable). One can show this. But this doesn't undermine the delimiting thesis, because the classes of intrinsic and extrinsic value are not mutually exclusive. A property can be intrinsically valuable and also be such that many or all of its instantiations bring about good consequences. Properties that are intrinsically valuable may be extrinsically valuable as well. To falsify the delimiting thesis, one would have to point to examples of variably relevant properties. It does seem easy to do this. Kicking someone's shins is impermissible in the schoolyard, acceptable in
Shafer-Landau
Moral Rules
593
a karate ring. Shaking hands is morally neutral when first meeting someone, morally significant when consummating a promise. So we cannot frame a moral rule that has kicking or handshaking as a grounding property. This is what we would expect; there are no such rules because such activities are not invariably (dis)valuable. More important, when they are (dis)valuable, they are so only because they instantiate some other, morally relevant property (e.g., harming others, keeping promises). For all we've said, it may be true that these other, morally relevant properties do delimit the set of moral values. This allows us to clarify the way in which the delimiting thesis might be falsified. Providing examples of variably valuable properties isn't quite enough, as the examples of kicking and handshaking showed. Instead, the delimiting thesis is false if and only if there are morally relevant properties whose different instantiations exhibit different (presumptive) moral value. A property is morally relevant just in case the (presumptive) moral value of an action or trait is most directly explained by the property's being instantiated. Handshaking per se isn't morally relevant; it generates certain moral expectations and obligations when and because it is an instance of promise keeping. It's not the kicking, but the harm that the kicking brings about, that makes a bully's behavior condemnable. It must be morally relevant properties that are variably valuable in order to undermine the delimiting thesis. One thing is clear: we can't resolve the debate about the delimiting thesis by citing examples. Those with intuitions favoring the thesis will see some good in each case of, say, generosity, no matter the horrific consequences that might ensue. Those with competing intuitions will cite such cases as instances in which generosity's usual value is entirely absent, or reversed. We could resolve this debate if we had a plausible account of justifying assignments of intrinsic value, but we are still largely in the dark on that matter. Second Reply: Spectjication The delimiting thesis may be true even if we could show some cases of generosity utterly lacking in value. If generosity were often, but not always, a right-making feature of actions, then generositysimpliciter could not be a grounding property in a moral rule. But generosity might still be invariably relevant under certain specified conditions. If these conditions are enumerable, then hope would remain for the delimiting thesis. Generalists (those who endorse an ineliminable place for rules in ethics) would have to refashion their rules to append exceptive clauses to the general grounding properties. For example, instead of the rule "generosity is presumptively right," which is false if generosity is of variable relevance, the rule would have to be nar-
594
Ethics
July 1997
rowed to read, "generosity is (presumptively) right except in circumstances a, b, c. . . ."" The plausibility of this move cannot be decisively determined except through consideration of the candidate specified rules that are brought forward.12Appeals to the complexity of morality, the possibility of novel situations, the need for flexibility, improvisation, and context-sensitive appreciation of moral salience cannot be known in advance to undermine the generalist project. Indeed, generalists who are prepared to concede the variable value of broad properties like generosity can claim that it is precisely their appreciation of this complexity that underwrites their abandonment of the allegedly overgeneral, Rossian prima facie rules, and moves them to carve out narrower moral rules. Of course, if the rules become so narrow as to incorporate temporal indexicals, proper names, a referential use of definite descriptions, or other nonrepeatable properties, then we have a theory no different from particularism. But specificationismpromises repeatably instantiable grounding properties. The only way to tell whether this promise can be discharged is to examine the candidates seriatim. We cannot know in advance whether the casuistry of cases will yield irreducibly singular morally salient features, or whether all such features will be instances of uniformly relevant properties. Assume for purposes of argument that we do know that the broad Rossian grounding properties are variably relevant. One might still endorse the delimiting thesis, and hence welcome the move to specification, because one is drawn to a familiar picture of moral justification from which the delimiting thesis very naturally follows. The picture is one in which we justify moral attributions in particular cases only by reference to the repeatably relevant properties that are being instantiated there. For instance, the badness of, say, a killing is explained 11. For a defense of this kind of move, see Henry Richardson, "Specifying Norms as a Way to Resolve Concrete Ethical Problems," Philosophy and Public Affairs 19 (1991): 279-3 10. I provide a discussion of specification, with an emphasis much different from Richardson, in a festschrift volume for Joel Feinberg: see my "Specifying Absolute Rights," Arizona Law Review 37 (1995): 209-25. Strictly speaking, there are two sorts of specification that might be undertaken. We might (i) narrow the grounding property itself, or (ii) retain the generally described grounding property but append exceptions that enumerate the kinds of occasions on which it loses its ordinary force. Schematically, (i) If Glx or G2xor GSx, then Mx; (ii) If Gx, then Mx, except in circumstances C1, CP,or C3 12. Contrast Judith Jarvis Thomson, "Self-Defense and Rights," reprinted in her Rights, Restitution and Rkk (Cambridge, Mass.: Harvard University Press, 1986), pp. 33-48. Thomson seeks to preempt consideration of candidate rules by a set of general arguments against specification. The strongest criticisms are that (i) specified rules fail to be appropriately morally explanatory; (ii) compensatory claims cannot be justified by means of such rules; (iii) specified rules yield a convoluted moral epistemology; (iv) specification entails an odd moral ontology. I defend specificationism against each of these criticisms in "Specifying Absolute Rights."
Shafer-Landau
Moral Rules
595
by its being an instance of a kind of action that is invariably immoral. If we had to concede that the bad-making feature of this killing is different from that of any other, then (according to this account) we effectively abandon the possibility of rational moral explanation and justification. Adherents of the delimiting thesis must allow that people may exemplify unique traits and that these traits may assume moral relevance. They must allow that repeatable features can combine to form nonrepeatable morally laden situations. They should grant that certain things and all people are nonreplaceable. The generalist insists, however, that to describe what is valuable in each case is to invoke properties whose value is pervasive. On this account, explaining value would be impossible if we were to allow a feature to once assume moral importance, but deny it a similar role when exemplified elsewhere. The notion of sui generis moral relevance is allegedly mysterious and is thought to confer no explanatory power.13 Particularists, of course, will disagree about the need to invoke general properties to explain singular moral attributions. The particularist will explain the wrongness of some killing by citing the features of the killing and its circumstances. If puzzled about whether or why these features are relevant in the way particularists insist, the interlocutor must simply be invited to look again, carefully. The moral relevance of particular features is a brute fact,14 not explicable by reference to uniformly valuable properties. Particularist moral explanations are too brief if and only if there really are uniformly relevant properties that can be invoked to assist explanation. Thus we simply beg the question if we try to support the delimiting thesis by relying on the familiar generalist picture of moral justification. This conception presupposes the truth of the delimiting thesis; it assumes that for every morally relevant feature of the world, its relevance can be explained only by identifying it as an instance of a uniformly relevant property. Still, the generalist view does have deep intuitive appeal. It may be true. But it pays to see precisely how it is used in the anti-particularist argument:
13. Shelly Kagan appreciates the force of the relevant generalist intuition here, all the while remaining suspicious of what I am calling the delimiting thesis. For a fine, complementary discussion of many of the issues I raise in this article, see Kagan, "The Additive Fallacy,"Ethics 99 (1988): 5-31. 14. No commitment to realism or other forms of objectivism is intended by the reference to brute facts. I mean all references to such facts to be construed according to the reader's metaphysical preferences.
596
Ethics July1997 1. If something is valuable, its value is either a brute fact, or its value is explicable as being an instance of an invariably relevant property. 2. Appeal to brute facts vitiates the explanatory power of a theory that contains such an appeal. 3. Particularists must often (or always, if strong particularists) appeal to brute moral facts. 4. Therefore particularism, to the extent that it must appeal to brute facts, lacks explanatory power.
Suppose we grant that the kind of explanatory failure that comes from failing to cite uniformly valuable properties is fatal for a theory. The problem is that generalism would be undermined as well. If premise (1) is true, then the generalist must have recourse to brute moral facts: for some uniformly relevant properties, there is no more general property from which to deduce their relevance. This must be so if we are to avoid an infinite explanatory regress. Thus the generalist must identify the value of some invariably relevant property or properties as a brute fact. If premise (2) is true, it follows that generalism lacks explanatory power. Premise (1) may be false. We might be able to justify assignments of value without invoking more general properties. But absent any argument, we must assume that whatever method of justification is offered is equally available to the particularist, in which case particularists would not be saddled with having to rely on brute facts. If premise (1)is true, we could save generalism by rejecting premise (2), the claim that reliance on brute facts vitiates explanatory power. But then we'd lack any reason for rejecting particularism. Generalists might profit from this argument only if they add two additional premises: 5. Other things equal, theory A is more likely than B to be true if A contains fewer citations of brute facts. 6. Generalist theories will contain fewer citations of brute facts than particularist theories. Endorsing these premises entails that reliance on brute facts isn't by itself sufficient to undermine a theory; particularism wouldn't be undermined just because it contains references to brute facts, but because it contains a greater number of such references than generalist theories do. Premise (6) is probably true. But as yet we've no reason to suppose that the ceteris paribus clause of (5) is satisfied in the context of the generalism-particularism debate. Further, a generalism that endorses the move to specification may contain hundreds, perhaps thousands, of intrinsically valuable properties with their specified exceptions. For all we know, most of these would be underivable from any broader invariably morally relevant property. So even if (5) and (6) were true, the possibly very high number of brute facts to which
Shafer-Landau
Moral Rules
597
a generalist need appeal shows that the argument from (1)-(6) cannot carry much weight. We've yet to receive anything like a convincing rebuttal of moderate particularism (the sort that allows for prima facie rules but rejects the delimiting thesis). Alternatively, moderate particularists have nothing but a promissory note to offer by way of criticizing the delimiting thesis. The note is cashable if and only if they can adduce examples containing features whose moral relevance cannot be explained by reference to any repeatably instantiable, invariably relevant property. As it stands, the case against the delimiting thesis cannot proceed until we have a defensible theory enabling us to identify intrinsic value. Suppose we had such an account. The delimiting thesis would be false only if (i) correctly applying this account shows that some of the properties we take to be invariably morally relevant are not, and (ii) no suitably narrow specification of these properties is uniformly valuable. V. FIXING MORAL STATUS: PRIMA FACIE RULES Even if the delimiting thesis is true, this leaves us far short of a fixing role for moral rules. To explore this role, we need to distinguish three possible generalist paths in which rules, in conjunction with specific case descriptions, might fix moral conclusions. Though rarely discussed, it is at least possible to see certain strategies for prima facie rules as capable of yielding determinate conclusory moral verdicts. The alternative is to embrace a form of absolutism. Here the choice is between different kinds of monism (that identify just a single absolute rule) and different types of pluralism (that allow for many such rules). If the rules that can identify (and possibly delimit) morally relevant features are merely prima facie, then moral rules could supply conclusory moral verdicts in those cases where just a single prima facie rule applies, or those cases where all of the relevant prima facie rules concur in their recommendation. But what of the cases in which prima facie rules point in conflicting directions? Traditionally, prima facie rules are here thought incapable of fixing a moral verdict. The assumption underlying the traditional view is that the weights of prima facie rules can alter depending on circumstances. Since we are unable in advance to fix the weight of each competing prima facie consideration, we must go beyond the rules themselves-to intuition or to some absolute rule-to provide a resolution in cases of conflict. If intuitionism and absolutism are misguided, then there is no way to nonarbitrarily adjudicate such conflicts. Indeterminacy would be rife in the moral domain. We can stave off widespread indeterminacy either by defending intuitionism or by showing some form of absolutism justified. Or we could challenge the underlying assumption of the traditional view. If
Ethics
598
July 1997
there were an invariant weighting of all prima facie rules, then these weights would provide a way of fixing moral verdicts without invoking absolute rules or an intuition that could detect which of many variable weights a prima facie consideration possessed in a given case. A rank ordering that establishes once and for all a hierarchy of prima facie considerations would be either ordinal or cardinal. Suppose the ordering is provided by an exhaustive ordinal ranking of all the prima facie rules. The problem with this model is that it fails to show how a lower-ranked prima facie rule can ever outweigh a higher-ranked one. It will also fail to yield any resolution in cases where a prima facie rule conflicts with a plurality of lower-ranked rules. This last is a feature of any merely ordinal ranking, and so an ordinal ranking will do little to alleviate worries of pervasive ethical indeterminacy. The alternative is to assign cardinal weights to the prima facie rules. This will solve both problems encountered by the ordinal model. For instance, if (fancifully) telling the truth = 6, and keeping promises 5, then even though truth telling always counts for more, we can envision scenarios where we are required to keep a promise even if we have to lie to do so. This is possible because an act of promise keeping may be enjoined by many rules whose total weight overrides the total associated with those rules that in this particular context support truth telling. But different circumstances may yield a different outcome. Unlike the method of ordinal ranking, this method shows both how we can allow lower-ranked requirements to sometimes override higher ones and how we can combine the edicts of a plurality of rules. This model will surely minimize moral indeterminacy, but it won't eliminate it unless prima facie rules succeed in delimiting the class of morally relevant features. For if some morally relevant features of situations escaped incorporation into rules-if some morally relevant properties were only variably valuable-then the relevance they bring to a case would be omitted from a calculus that takes only the cardinal weights of prima facie rules into account. Perhaps this is something we could live with, since many think that we should leave room for pockets of indeterminacy in morality. But we get this far in solving conflicts only if we grant the plausibility of the model of cardinal weights. This model strikes me as entirely unbelievable. For starters, there is no plausible account of what the cardinal units would represent. And, more damaging, a cardinal model presupposes the truth of some fixed scalar ranking of prima facie rules. But it seems fantastic to suppose that the intrinsic merit of each specified kind of generosity must be either permanently greater or less than that of each specified kind of honesty, truth telling, orjustice. I don't know of a way to prove the impossibility of a fixed cardinal ordering of all prima facie rules. But if, as I suggest, such an ordering is deeply implausible, then we must return to the traditional view that
+
+
Shufer-Landau
Moral Rules
599
sees the grounding properties of prima facie rules as possessed of variable weights. If that is right, and if most morally laden cases are governed by conflicting prima facie rules, then prima facie rules will usually be incapable of fixing moral verdicts. We need to explore absolutist and intuitionist theories if we are to avoid the almost complete moral indeterminacy that would otherwise result. VI. FIXING MORAL STATUS: MONISTIC ABSOLUTISM The absolutist thesis can be put as follows: there is at least one invariably determinative grounding property. A property is invariably determinative if and only if its every instantiation fixes the moral status of the situation in which it appears. Rule monists believe that there is one and only one morally determinative grounding property. So there is just a single absolute moral rule. There are as many monistic theories as there are views about which grounding properties are mere pretenders, and which the real thing. The economy of a rule monism is maximized if its axiology is monistic as well. For such theories, there is just a single kind of fundamental value, and the single absolute moral rule makes rightness some function (usually a maximizing one) of goodness. Any such theory is implausible, however, since the value monism that it presupposes is almost certainly false. Certain kinds of pleasure, knowledge, autonomy, and family love are each valuable for their own sake. None of these values can be reduced to or derived from another of them. Efforts to identify just a single basic value typically falsify the true scope of the realm of value for the sake of theoretical simplicity. They either blind themselves to the real worth of certain experiences or activities or ultimately enlarge the nature of the alleged single value to such an extent as to make the theory pluralistic in all but name. This last option recalls Mill's efforts to save his hedonism by arguing that virtue and knowledge are really just different species of pleasure. For every sort of experience or activity that appears to be intrinsically valuable and distinct from pleasure, we simply claim it to be a constituent of pleasure. But saying it is one thing, making it so another. Knowledge is not pleasure, or a form of pleasure, and does not derive what value it has entirely from the pleasure it occasions. There is no plausible candidate supervalue to which all others can be reduced. The success of rule monism thus hinges on its ability to identify a rule that can (i) acknowledge a plurality of irreducible values and (ii) avoid commitment to an invariant ranking of value types while (iii) offering a method of determinately ranking instances of these values in different contexts. This won't be easy. In my opinion, the most promising monistic option here is to adopt some sort of constructivism. What makes an action right, on any such account, is its eliciting the appropriate response from an
600
Ethics
July 1997
agent (or agents) whose choices are determinative of right conduct. So, for instance, an act is right if (and only if) it (i) elicits approval from a duly specified ideal observer, (ii) conforms to a rule no one could reasonably reject as the basis of informed, unforced general agreement, or (iii) would be desired as choiceworthy by oneself if possessed of relatively full information. l5 The differences among these theories needn't concern us here. The point is that, for each of them, an agent or group of agents, under specified conditions of choice, fix the moral verdicts by their responses. Only a certain kind of constructivism will assist the rule monist, however. Those constructivisms whose verdict-fixing agents employ rules to parcel out right and wrong, or whose choices can be reconstructed as a set of moral rules, are tantamount to rule-pluralistic theories, and so will be of little use to the rule monist. For instance, in Rawls's work, or Scanlon's, the objects of constructed choice are sets of rules.16 In discovering or reconstructing what the appropriate agents of choice will decide upon, we arrive at a set of rules which we then take to be determinative of justified moral or politicaljudgments. Thus the kind of constructivism that can be serviceable to the rule monist is one in which the appropriate agents of choice neither proceed with rules in hand, nor make choices from which rules can be inferred. In this case, we really do have just a single rule-do what the (idealized) agents would recommend-and we can't know what they'd recommend by trying to identify a battery of rules that they would employ or generate. Firth's and Griffin's theories are of this kind.17 The prospects for rule monism hinge on whether a general problem for constructivism,and a special problem for the monistic versions, can be disposed of. The general problem is this. We enshrine the responses of the agents of construction as the appropriate means of fixing moral conclusions. But why should we take the responses of these agents as constitutive of correct ethical assessments? (Perhaps they are merely good evidentiary guides for correct ethical choice.) Consider Griffin's theory. Choice X is morally more important than Y because I prefer X, given adequate information about my choices and situation. Either I can give reasons for my preference or I cannot. A reasonless 15. For (i), see Roderick Firth, "EthicalAbsolutism and the Ideal Observer,"Philosophy and Phenomenological Research 12 (1952): 317-45; and Thomas Carson, The Status of Morality (Dordrecht: Reidel, 1984). For (ii), see Thomas Scanlon, "Contractualism and Utilitarianism,"in Utilitarianism and Beyond, ed. Bernard Williams and Amartya Sen (Cambridge: Cambridge University Press, 1982), pp. 103-28. For (iii), see James Griffin, Well-Being (Oxford: Oxford University Press, 1985). 16. John Rawls, A Theory $Justice (Cambridge, Mass.: Harvard University Press, 1971), "Kantian Constructivism in Moral Theory," Journal of Philosophy 77 (1980): 515-72, and Political Liberalism (New York: Columbia University Press, 1993); Scanlon. 17. Firth; and Griffin.
Shafer-Landau
Moral Rules
60 1
preference hardly seems the sort of thing to fix moral conclusions. On the other hand, a reason-based conclusion lets us substitute those reasons as the justifying grounds for the conclusion, rather than my desires. My desires fall away as epiphenomena, doing no real justificatory work. The same holds for any process of construction. It is the grounds on which the agents of construction choose, rather than their responses, that fix moral verdicts. Suppose this challenge could be answered. Still, once we allow that the idealized agents will not reach their decisions by applying rules to cases, don't we effectively give up the ship in favor of particularism? Its weakest form claims that there are no features of the world such that the presence of any one is invariably sufficient to determine a moral verdict. This is precisely what is admitted by rule monists who have their idealized agents respond to cases without the use of absolute rules. If idealized agents fail to identify always relevant and determinative grounding properties in the cases they survey, then there are no invariably determinative moral rules. There arejust the unsystematizable responses of idealized agents. This abandonment of rules on the part of idealized agents and the correlative rejection of uniformly determinative properties surely give weak particularists all they might want. VII. FIXING MORAL STATUS: PLURALISTIC ABSOLUTISM The Argument from Moral Horror A central feature of absolute rules is that their grounding properties, if once determinative of a verdict, must always be so (otherwise the rule would be consigned to prima facie status). I have called these grounding properties "invariably determinative properties." If there are any such, then some moral rules are absolute. A familiar way to reject the existence of absolute rules is to insist that whatever their content, they are bound to conflict with one another, thereby undermining their absolute status. Rather than pursue this line of criticism, which seems weakened once we appreciate the possibility of specifying grounding properties, I'd prefer to examine another traditional argument against absolutism. Call this the "argument from moral horror": 1. If moral absolutism is correct, then we may be, all things considered, morally required to do or forbear from certain activities even though such action or omission generates horrific consequences. 2. Morality does not require such actions or omissions. 3. Therefore moral absolutism is false. Even Nozick, a staunch absolutist, allows that cases of "catastrophic moral horror" may require suspension of absolute side constraint^.'^ 18. Robert Nozick, Anarchy, State and Utopia (New York: Basic, 1974), p. 30n.
602
Ethics
July1997
Attention to the dire consequences that may be brought about by allegiance to absolute rules needn't move us to the consequentialist camp-it didn't incline Ross or Nozick in that direction, for instance. But it does create a presumptive case against absolutism. Absolutist responses to the argument standardly take one of two forms. The first is to reject premise (1) and deny that absolutism generates tragic consequences, by arguing that a set of suitably narrowed absolutist rules will not require behavior that results in "catastrophic moral horror." The second response is to reject premise (2) and defend the moral necessity of obedience even if tragic consequences ensue. Rejecting Premise ( I ) Consider the first strategy. This is tantamount to a specificationist program that begins by admitting that the standard candidates-don't kill, lie, cheat, commit adultery-cannot plausibly be construed as absolute rules. Just as we had to narrow their scope if we were to show them universally relevant, so too we need to narrow the scope of such properties to show them universally determinative. The question, though, is how far, and in what way, this added concreteness is to be pursued. The double dangers that the absolutist must avoid at this juncture are those of drawing the grounding properties too broadly, or too narrowly. Rules drawn too narrowly will incorporate concrete details of cases in the description of the grounding properties, yielding a theory that is particularist in all but name. The opposite problem is realized when we allow the grounding properties to be drawn broadly enough as to be repeatably instantiated, but at the cost of allowing the emerging rules to conflict. Some middle ground must be secured. How could we frame an absolute rule that enjoinedjust the actions we want, while offering an escape clause for tragic cases? There seems to be no way to do this other than by appending a proviso to the rule, to the effect that it binds except where such obedience will lead to catastrophic consequences, very serious harm, horrific results. Because of the great variety of ways in which such results can occur, there doesn't seem to be any more precise way to specify the exceptive clause without reducing it to an indefinitely long string of too-finely described scenarios. Is this problematic? Consider an analogous case. Someone wants to lose weight and wants to know how long to maintain a new diet. A dietician offers the following advice: "Cut twenty percent of your caloric intake; this will make you thinner, but also weaker. If you reach a point where you've gotten too thin and weak, increase your calories." The dietician's advice is flawed because it doesn't give, by itself, enough information to the person trying to follow it. It's too general. The
Shafer-Landau
Moral Rules
603
qualified moral rule is similarly uninformative. If abiding by the rule will occasion harmful results, one wants to know how harmful they have to be to qualify as too harmful. The rule doesn't really say-'catastrophic' is just a synonym for 'too harmful'. Such a rule is crucially underspecific, and this undermines efforts to apply it as a major premise in deductive moral argument. This lack of specificity results from an absence of necessary and sufficient conditions that could determine the extension of the concept "catastrophic conseq~ences."'~ Efforts to remove this underspecificity by providing a set of definitional criteria typically serve only to falsify the resulting ethical assessments; imagine the futility of trying to precisely set out in advance what is to count as catastrophic consequences. Rendering the notion of "catastrophic" more precise seems bound to yield a rule that omits warranted exceptions. Or it may cover all such exceptions, but at the cost of making the exceptive clause so finegrained that it will be nothing less than an indefinitely long disjunction of descriptions of actual cases that represent exceptions to the general rule. Neither option should leave us very sanguine about the prospects of specifying absolute rules so as to ensure that such rules can be obeyed without occasioning catastrophic consequences. Rejecting Premise (2) The alternative for the absolutist is to stand fast and allow that morality requires adherence to rules that will sometimes yield catastrophic horrors. There is no inconsistency in taking such a stand. But the ethic that requires conduct that is tantamount to failure to prevent catastrophe is surely suspect. Preventing catastrophe is presumptively obligatory. The obligation might be defeasible, but absolutists have yet to tell the convincing story that would override the presumption. Imagine that you are a sharpshooter in a position to kill a terrorist who is credibly threatening to detonate a bomb that will kill thousands. If you merely wound him, he will be able to trigger the firing mechanism. You must kill him to save the innocents. Suppose that in obedi-
19. Not that providing definitional criteria is any guarantee that underspecificity will be eliminated. Though we can define 'bravery', for instance, this isn't sufficient to fix everything that belongs within and without the set of brave things. One is brave if and only if one (i) faces significant danger, (ii) appreciates the nature of the threatened harm, and (iii) stands fast before it, (iv) provided the danger is not too great. But what's to count as "significant," as "too greatn-even as "standing fastM-is not itself determined by the rule. Nor is it given by a further set of rules. (What would such rules look like?) These notions, crucial to an understanding of bravery, are themselves underspecific. To know whether someone is brave, or some consequences catastrophic, we need to exercise a kind of judgment that is something very different from trying to apply rules to cases.
604
Ethics
July 1997
ence to an absolutist ethic you refrain from shooting. The terrorist detonates the bomb. Thousands die. Something must be said about the agent whose obedience to absolute rules occasions catastrophe. It is possible that an absolutist ethic will blame you for doing your duty. Possible, but unlikely. Absolutists who allow that obedience to their favored rules may occasion catastrophe typically seek ways to exculpate those whose obedience yields tragic results. The standard strategy is to endorse some version of the doctrine of double effect, or the doctrine of doing and allowing. The former says that harms brought about by indirect intention may be permissible even though similar harms brought about by direct intention are forbidden. The latter says that bringing about harm through omission or inaction may be permissible even though similar harms brought about by positive action are forbidden. The motivating spirit behind both doctrines is to legitimate certain kinds of harmful conduct, to exculpate certain harm doers, and to forestall the possibility that absolute rules might conflict. The truth of either doctrine would ensure that agents always have a permissible option to pursue-namely, obedience to an absolute moral rule.*' Quite apart from the fact that these doctrines have yet to be adequately defended,*' their adequate defense would still leave us short
20. If failure to obey an absolute rule is always wrong, and absolute rules can conflict, then genuinely tragic cases of moral dilemma may arise. To avoid this possibility, absolutists-needto show either that (i) no such conflict can arise, or (ii) violation of an absolute rule is sometimes permissible. This last has been thought implausible, and so we have a ready explanation of the absolutists' desire to avoid conflict among their favored rules. This explanation forces a particular reading of what it is for a rule to be maximally stringent. This notion is ambiguous between (1) never overridden and (2) always overriding. The stronger (2) is necessary if absolute rules are to fix moral verdicts. Option (1) would allow conflict among rules that would yield moral dilemma. (Walter Sinott-Armstrong [Moral Dilemmus (Oxford: Basil Blackwell, 1988)],seeking to defend the existence of moral dilemmas, endorses [I] but not [2].) To serve a fixing role, the presence of a grounding property in such a rule must be determinative of a verdict. If (1) alone is the proper reading of maximal stringency, then so-called absolute rules can conflict and will no longer be always determinative. Only by thinking of absolute rules as never overridden and always overriding will one ensure a fixing role for such rules, by eliminating the prospects of conflict. 21. The general problem for advocates of double effect is their inability to identify a conceptuallycoherent account of direct and indirect intention that also bears the moral weight needed to make distinctions among their paradigm cases. The most sophisticated defense of the doctrine is presented by Warren Quinn, "Actions, Intentions and Consequences: The Doctrine of Double ~ffect,"philosophy and Public Affairs 10 (1989): 334- 5 1. It is ably criticized in a wide-ranging article by Donald Marquis, "Four Versions of Double Effect," Journal of Medicine and Philosophy 16 (1991): 515-44. A similar problem arises for proponents of the doctrine of doing and allowing. Kant and Mill, and many others following in their footsteps, tried to provide a distinction between perfect and imperfect duties that would be both conceptually plausible and morally
Shafer-Landau
Moral Rules
605
of a justification of the absolute rules that are to complement them. Neither of these doctrines is itself a defense of absolutism; rather, they are really "helping doctrines," whose truth would undermine the inevitability of conflict among absolute rules. We may always have a permissible option in cases where we must choose between killing and letting die, intending death or merely foreseeing it, but this by itself is no argument for thinking that the prohibition on intentionally killing innocents is absolute. If these familiar exculpatory doctrines are unjustified, then we have reason to retain our initial inclination to fault those whose conduct knowingly occasions catastrophe. This doesn't prove absolutism false. It might sometimes be true that agents deserve blame for doing their duty. It might be true that agents deserve blame even though they've done their duty while prompted by a sense of duty. But how much simpler to suppose that their duty lies elsewhere, that an ethic requiring conduct that yields catastrophic horrors is mistaken.
Coverage On the assumption that there are some determinate moral verdicts, one of the following must be true: ( F ) every determinate moral verdict is fixed by an absolute rule, or ( - F ) a t least one determinate moral verdict is not fixed by an absolute rule. If (F) is true, then every determinative grounding property must be invariably determinative. Every property that once fixes a verdict would fix verdicts in every situation in which it is instantiated. If ( F ) is false, then some properties that are once determinative will fail to be determinative in other contexts. Because such properties do not invariably ground conclusory moral verdicts, these properties cannot properly be located in an absolute rule. The argument from moral horror gave us reason to believe that there are no absolute moral rules. Assume now that this argument
significant. But all efforts to define the separate spheres have failed to mark an invariably morally relevant distinction. The standard strategy is to divide requirements into a set of negative and positive ones, elevating the negative to absolute or perfect status, while allowing the positive obligations to be overridable. But no one has provided a conceptually plausible account of the negative and the positive duties that also justifies (i)the-assumptionthat negative requirements cannot conflict, and (ii) the absolute moral priority of negative requirements over positive ones. Quinn ("Actions, Intentions and Consequences: The Doctrine of Doing and Allowing," Philosophical Review 98 [1989]: 287-312) has provided the most sophisticated defense of the doctrine of doing and allowing. It has been soundly criticized by John M. Fischer and Mark Ravizza, "Quinn on Doing and Allowing," Philosophical Review 101 (1992): 343-52.
606
Ethics
July 1997
has been defeated. Assume further that we have overriding reason to believe that there are some absolute moral rules. The question naturally arises as to the extent of cases for which determinate verdicts are made available by such rules. Is every determinative property invariably determinative? What is at issue here is really another kind of delimiting thesis(F)-that claims that absolute rules are capable of delimiting the domain of moral verdicts; absolute rules, in conjunction with specific case descriptions, can fix all determinate conclusory moral assessments. Thesis ( F ) is true if and only if every morally determinative property is invariably determinative; ( F ) also entails that every property that is once instantiated but fails to be determinative may never be determinative. So if ( F ) is true, then every morally relevant property must be either never morally determinative or always morally determinative. Note that ( F ) is incompatible with a model of prima facie rules. The whole point of such a model is to allow the moral importance of a grounding property to sometimes override another and to sometimes be overridden. In short, what distinguishes a prima facie rule from an absolute rule is that the grounding properties of the former are variably determinative and those of the latter cannot be. If ( F ) is true, there are no variably determinative properties. Thus if ( F ) is true, there are no prima facie rules. A complete battery of absolute rules would force us to eliminate prima facie rules, or retain them as mere epistemic guides. Since it is initially plausible to suppose that there are some prima facie rules, it is initially plausible to assume the soundness of the following argument: 1 . If ( F ) is true, then there are no variably determinative properties. 2 . Some properties are variably determinative. 3. Therefore ( F ) is false. Premise ( 1 ) is an analytic truth. If the argument from moral horror is sound, then premise ( 2 ) is true as well, since the soundness of that argument would show that every determinative property is variably determinative. Rather than rely on the soundness of that argument, we might instead support premise ( 2 )by citing the examples of invariably relevant properties I identified in Section 111. That an action is a cruel one is often sufficient to show it is wrong, though there are exceptions. Giving most of one's accumulated wealth to charity often decisively inclines us to an attribution of generosity, but not always. Taking pleasure at another's pain usually fixes a verdict of blameworthiness, though there are exceptions. Killing a child is presumptively wrong, but not always impermissible, all-things-considered.
Shufe-Landau
Moral Rules
607
These examples conclusively support (2) only if they are conclusive examples of variably determinative properties. They are not. Those who endorse ( F ) will claim that these properties are in fact never determinati~e.~~ It would be strange to view such properties as invariably relevant but never determinative. If these properties are never determinative, it must be because they aren't really invariably relevant. Enter the familiar move of insisting that we have mistaken an overgeneral, variably relevant counterfeit for its correlative invariably relevant specified property. Once we carve out the specified property, we would see that wherever it appears, it is determinative, contrary to first impression. To sustain this move, advocates of ( F ) must show not only that the candidate grounding properties I identified above are never morally relevant, but also that their substitutes are invariably relevant and invariably determinative. Perhaps the proponent of (F) could show this. But surely the burden of proof is on those who would explain away the appearances and in so doing argue that there can be no prima facie rules. This debate recalls that between the particularist and specificationist regarding the delimiting thesis. But while earlier specificationists could appeal to a deeply ingrained rationale for their position-the view that explaining moral salience requires that grounding properties be repeatably, uniformly relevant-no such rationale is forthcoming for ( F ) . Not only have absolutists failed to provide anything like an exhaustive battery of absolute rules, they have not even attempted to justify the claim that all determinative moral properties are invariably determinative. And there is nothing in the structure of absolutist theories that would require them to do so. Absolutists have always been intent on proving that there are some absolute rules, that some grounding properties are universally determinative, rather than the far stronger claim that variably determinative grounding properties cannot exist. As far as I know, there is no argument in the literature that supports ( F ) .Further, the theoretical underpinnings of absolutist theories entail no commitment to ( F ) .The examples offered above provide presumptive evidence against ( F ) .Given this, we have reason to follow the appearances and so accept the view that some grounding properties are variably morally determinative. If this is so, then absolute rules will, at best, fix verdicts in only a subset of determinate morally laden cases. At worst, they will fix none; if the argument from moral horror is sound, then there are no invariably determinative grounding properties.
22. They could claim that the properties are always determinative, but I will assume that this is implausible.
608
Ethics
July 1997
VIII. CONCLUSION, AND SOME IMPLICATIONS At this point, it might be helpful to review the spectrum of theses that we've considered thus far: 1. No properties are invariably relevant (strong particularism).
2. Some properties are invariably relevant, others are variably relevant (moderate particularism). 3. All properties that are once relevant are invariably relevant (the delimiting thesis). 4. No properties are invariably determinative (prima facie modellweak particularism). 5. Some properties are invariably determinative, others variably determinative (moderate absolutism). 6. All properties that are once determinative are invariably determinative (absolutism that endorses [F]). I have argued in Section I11 that thesis (1) is false. Sections V-VII can be seen as an extended argument against thesis (6). So far as I can tell, we have no grounds for believing thesis (6) true, and some reason, though perhaps not conclusive reason, to think it false. The choice between (2) and (3) cannot be resolved until we possess a developed method for determining whether, in a given case, a property's invariable, uniform relevance is being overridden, or whether its relevance has altered because of context. A complicating factor is the generalist possibility of making sound specificationist moves, in which a putative instance of a variably relevant property is more concretely described, thus rendering it uniformly and universally relevant. Locating the burden of proof in this debate is difficult. On the one hand, the delimiting thesis implies a position that assigns intrinsic value to (almost) every morally relevant property. That seems exceedingly strong. But it is too strong only if moderate particularism is true. So we can't use this consideration to tip the burden of proof in either direction. We might presumptively oppose moderate particularism because it is incompatible with a view of moral justification that has deep intuitive appeal. Yet we saw that this picture either undermines the delimiting thesis as well, or is neutral between it and moderate particularism. The argument from moral horror, and the discussion of the two absolutist responses, were designed to show that there is a presumption in favor of thesis (4) and against thesis (5). This presumption would be overridden by a sound theoretical account of the necessity of an exceptionless, always overriding moral rule. The presumption would be neutralized by providing a candidate property that intuitively appears to be invariably determinative and that survives counterexamples. I remain skeptical that either alternative has been successfully made out, but a full accounting cannot take place short of patient and
Shafer-Landau
Moral Rules
609
thorough investigation of absolutist theories or a sound argument showing the impossibility of such theoriesz3 One's choice of theses has important implications for moral epistemology and ontology. If thesis (6) is false, for instance, then we must align a moral epistemology with the fact that some moral verdicts are not deducible from moral rules. Let us inelegantly call such verdicts "undeducible moral facts." (As with brute facts, no metaphysical commitments are implied.) If justified conclusory moral belief requires citation of a covering rule, then we must suspend judgment about every case whose verdict is fixed by a variably determinative property. (If thesis [5] is false, then every moral conclusion is an undeducible moral fact. The conjunction of [5] with the requirement that we cite a covering rule would entail a thoroughgoing moral skepticism.) How can we have epistemic access to undeducible moral facts? Foundationalism is one natural answer. According to foundationalists, a belief is justified if and only if it is either directly, noninferentially justified, or derived by appropriate methods from directly justified beliefs. Beliefs stating undeducible moral facts will not satisfy any of the traditional criteria for directlyjustified belief: they are not analytic, infallible, incorrigible, or indubitable. And it is very unlikely that such beliefs are appropriately derivable from directly justified beliefs. First, if there are no directly justified moral beliefs, then worries about naturalistic fallacies must be dispelled before entertaining the possibility of mediately justifying beliefs stating undeducible moral facts. Second, even if some moral beliefs are directlyjustified, one cannot deduce a belief stating an undeducible moral fact from such a source. So the only foundationalist hope is to argue that there are directly justified moral beliefs, and that undeducible moral facts can be inductively derived from them. But this will be extremely hard (perhaps impossible) to show. We are faced with an undeducible moral fact just when a variably determinative property is instantiated. Such properties are specially bad bases for inductive inference, precisely because their moral weight (and possibly their relevance) alters in different circumstances. So foundationalism, at least as it is traditionally understood, is unlikely to assist us in justifying claims to have identified undeducible moral facts. Coherentism will fare no better, since coherentists, too, justify particular moral judgments only by subsuming them under rules. Their difference with foundationalists is not about what consti23. I take it that Samuel Scheffler's arguments against agent-centered restrictions imply the unsoundness of all absolutist theories. But I cannot comment here on the success of his arguments. See Scheffler, The Rejection of Consequentialism (Oxford:Oxford University Press, 1982), and "Agent-CenteredRestrictions, Rationality and the Virtues," Mind 94 (1985): 409-19.
6 10
Ethics
July 1997
tutes appropriate inference rules, but about the ways in which the justification-conferring rules are justified (namely, by coherence, not direct justification). If, with coherentists, we assume that all justification is inferential, then the foundationalist problems of mediatelyjustifying beliefs stating undeducible facts will beset coherentism as well. This doesn't undermine either foundationalism or coherentism, since we have not shown that we can be epistemicallyjustified in our claims to have identified undeducible moral facts. However, if we want to avoid skepticism on this front, we must turn from traditional theories and endorse some new kind of moral epistemology. My thought is that only an externalist, reliabilist intuitionism will do the trick. Judgments about cases not derivable from rules can bejustified directly (i.e., noninferentially), without claiming that they are self-evident or analytic or indubitable. Intuitive judgment may be defeasible. The intuitions we have about cases involving undeducible moral facts will be justified just in case they emerge from an appreciative process that reliably yields correct ethical judgments. Moral justification wouldn't essentially involve the correct grasp of moral rules, but rather the development of the appropriate kinds of sensitivities. This of course is just handwaving; the detailed arguments must be offered e l s e ~ h e r e . ~ ~ The debates about appropriate roles for moral rules also have direct implications for moral ontology. For instance, rejection of thesis (1)allows us to eliminate moral nominalism as a candidate moral ontology. (That we have reason to endorse the existence of some moral "universals" does not, of course, settle the issue about whether these types are to be construed realistically.) And the issues raised in the discussions surrounding specificationist alternatives may fruitfully be seen in the context of a larger ontological debate about the individuation of properties. If there is a sound argument favoring ontological parsimony, then we have some reason for viewing uniformly relevant properties as Ross did, rather than as specificationists would have us do. Specificationists will individuate their moral properties much more finely, and so be more expansive in their ontological commitments. Issues about the extent of moral indeterminacy are also at play in the debates about the proper role for moral rules. Moral indeterminacy arises whenever a moral judgment is neither true nor false, or a true judgment reveals that no option in a given comparison most exemplifies a particular moral property. Moral indeterminacy is ordinarily explained precisely as thesis (6) would explain it-situations create indeterminacies if they fail to be covered by an absolute rule. However, if we reject thesis (6), then the standard account is false; if (6) is false,
24. I provide the details of such a theory in "Ethical Intuitionism Redux" (unpublished manuscript).
Shafer-Landau
Moral Rules
61 1
there is at least one variably determinative property, and so at least one determinate moral verdict that isn't derivable from an absolute rule. In fact, I think we do need a much more sophisticated account of the nature and sources of moral indeterminacy than the ordinary one, and showing (6) false would provide the needed incentive for such an inve~tigation.~~ These and other issues in moral epistemology and ontology are naturally linked with conceptions about the proper theoretical role@) for moral rules. What I have done here is to begin a sketch of some implications that might be developed from the earlier arguments. The earlier arguments themselves are properly seen as prolegomena to a full accounting of the proper roles for ethical rules. I haven't tried to demonstrate the impossibility of absolutism, for instance, but instead have tried to show what philosophical work must be done if it is to be vindicated. Likewise, my aim has not been to prove or disprove the delimiting thesis, but to show how hard it is to do either. I hope that these efforts, inconclusive as they may be, have shown that common talk of moral rules must be replaced by a more nuanced appreciation of what rules are, and what roles they might play in ethics.
25. 1 have offered an account of the nature and sources of moral indeterminacy in "Ethical Disagreement, Ethical Objectivity and Moral Indeterminacy,"Philosophy and Phenomenological Research 54 (1994): 331 -45, and in "Vagueness,Borderline Cases and Moral Realism,"American Philosophical Quarterly 32 (1995): 83-96.
http://www.jstor.org
LINKED CITATIONS - Page 1 of 3 -
You have printed the following article: Moral Rules Russ Shafer-Landau Ethics, Vol. 107, No. 4. (Jul., 1997), pp. 584-611. Stable URL: http://links.jstor.org/sici?sici=0014-1704%28199707%29107%3A4%3C584%3AMR%3E2.0.CO%3B2-B
This article references the following linked citations. If you are trying to access articles from an off-campus location, you may be required to first logon via your library web site to access JSTOR. Please visit your library's website or contact a librarian to learn about options for remote access to JSTOR.
[Footnotes] 6
On Moral Properties Jonathan Dancy Mind, New Series, Vol. 90, No. 359. (Jul., 1981), pp. 367-385. Stable URL: http://links.jstor.org/sici?sici=0026-4423%28198107%292%3A90%3A359%3C367%3AOMP%3E2.0.CO%3B2-L 6
Ethical Particularism and Morally Relevant Properties Jonathan Dancy Mind, New Series, Vol. 92, No. 368. (Oct., 1983), pp. 530-547. Stable URL: http://links.jstor.org/sici?sici=0026-4423%28198310%292%3A92%3A368%3C530%3AEPAMRP%3E2.0.CO%3B2-1 8
On Moral Properties Jonathan Dancy Mind, New Series, Vol. 90, No. 359. (Jul., 1981), pp. 367-385. Stable URL: http://links.jstor.org/sici?sici=0026-4423%28198107%292%3A90%3A359%3C367%3AOMP%3E2.0.CO%3B2-L 9
On Moral Properties Jonathan Dancy Mind, New Series, Vol. 90, No. 359. (Jul., 1981), pp. 367-385. Stable URL: http://links.jstor.org/sici?sici=0026-4423%28198107%292%3A90%3A359%3C367%3AOMP%3E2.0.CO%3B2-L
NOTE: The reference numbering from the original has been maintained in this citation list.
http://www.jstor.org
LINKED CITATIONS - Page 2 of 3 -
11
Specifying Norms as a Way to Resolve Concrete Ethical Problems Henry S. Richardson Philosophy and Public Affairs, Vol. 19, No. 4. (Autumn, 1990), pp. 279-310. Stable URL: http://links.jstor.org/sici?sici=0048-3915%28199023%2919%3A4%3C279%3ASNAAWT%3E2.0.CO%3B2-U 13
The Additive Fallacy Shelly Kagan Ethics, Vol. 99, No. 1. (Oct., 1988), pp. 5-31. Stable URL: http://links.jstor.org/sici?sici=0014-1704%28198810%2999%3A1%3C5%3ATAF%3E2.0.CO%3B2-0 15
Ethical Absolutism and the Ideal Observer Roderick Firth Philosophy and Phenomenological Research, Vol. 12, No. 3. (Mar., 1952), pp. 317-345. Stable URL: http://links.jstor.org/sici?sici=0031-8205%28195203%2912%3A3%3C317%3AEAATIO%3E2.0.CO%3B2-P 16
Kantian Constructivism in Moral Theory John Rawls The Journal of Philosophy, Vol. 77, No. 9. (Sep. 9, 1980), pp. 515-572. Stable URL: http://links.jstor.org/sici?sici=0022-362X%2819800909%2977%3A9%3C515%3AKCIMT%3E2.0.CO%3B2-1 21
Actions, Intentions, and Consequences: The Doctrine of Double Effect Warren S. Quinn Philosophy and Public Affairs, Vol. 18, No. 4. (Autumn, 1989), pp. 334-351. Stable URL: http://links.jstor.org/sici?sici=0048-3915%28198923%2918%3A4%3C334%3AAIACTD%3E2.0.CO%3B2-P 21
Actions, Intentions, and Consequences: The Doctrine of Doing and Allowing Warren S. Quinn The Philosophical Review, Vol. 98, No. 3. (Jul., 1989), pp. 287-312. Stable URL: http://links.jstor.org/sici?sici=0031-8108%28198907%2998%3A3%3C287%3AAIACTD%3E2.0.CO%3B2-R
NOTE: The reference numbering from the original has been maintained in this citation list.
http://www.jstor.org
LINKED CITATIONS - Page 3 of 3 -
21
Quinn on Doing and Allowing John Martin Fischer; Mark Ravizza The Philosophical Review, Vol. 101, No. 2. (Apr., 1992), pp. 343-352. Stable URL: http://links.jstor.org/sici?sici=0031-8108%28199204%29101%3A2%3C343%3AQODAA%3E2.0.CO%3B2-A 23
Agent-Centred Restrictions, Rationality, and the Virtues Samuel Scheffler Mind, New Series, Vol. 94, No. 375. (Jul., 1985), pp. 409-419. Stable URL: http://links.jstor.org/sici?sici=0026-4423%28198507%292%3A94%3A375%3C409%3AARRATV%3E2.0.CO%3B2-8 25
Ethical Disagreement, Ethical Objectivism and Moral Indeterminacy Russ Shafer-Landau Philosophy and Phenomenological Research, Vol. 54, No. 2. (Jun., 1994), pp. 331-344. Stable URL: http://links.jstor.org/sici?sici=0031-8205%28199406%2954%3A2%3C331%3AEDEOAM%3E2.0.CO%3B2-9
NOTE: The reference numbering from the original has been maintained in this citation list.
