TY - JOUR
AU - Miller, Dale, E
AB - Abstract Rule consequentialism holds that an action's moral standing depends on its relation to the moral code whose general adoption would have the best consequences. Heretofore rule consequentialists have understood the notion of a code's being generally adopted in terms of its being generally obeyed or, more commonly, its being generally accepted. I argue that these ways of understanding general adoption lead to unacceptable formulations of the theory. For instance, Brad Hooker, Michael Ridge, and Holly Smith have recently offered different answers to the question of what ‘acceptance rate’ we should build into our formulation of rule consequentialism, and all are unsatisfactory. I argue instead for a novel approach to formulating rule consequentialism, ‘uniform-moral-education’ rule consequentialism, on which what it means for a moral code to be generally adopted is not for it to be generally followed or generally accepted, but instead for it to be generally taught. rule consequentialism, moral education, Brad Hooker, Michael Ridge, Holly Smith I. INTRODUCTION Due mostly to the efforts of Richard Brandt, Brad Hooker, and Derek Parfit, rule consequentialism has become one of the ‘contenders’—the ethical theories generally regarded as worthy of serious discussion. The theory can be characterized in terms of two tenets: The moral standing of actions—whether they are obligatory, right, or wrong—depends on whether they are required, permitted, or forbidden by an ‘authoritative’ moral code or set of moral rules, and The authoritative moral code is the ‘ideal code’, the code whose general adoption would be optimal—that is, would have the highest expected value.1 Questions remain about how best to formulate rule consequentialism, including what it means for a moral code to be ‘generally adopted’. The usual view today is that adoption is best understood in terms of acceptance, and Brad Hooker, Michael Ridge, and Holly Smith have recently debated what ‘acceptance rate’ rule consequentialists should assume when calculating the expected value of a code's being generally accepted. Given the shortcomings of these acceptance-rate-based formulations of the theory, however, a different approach is needed. I will therefore propose a sea change in how we formulate rule consequentialism, one according to which what it means for a code to be generally adopted is for it to be generally taught. II. ADOPTION, COMPLIANCE, AND ACCEPTANCE It might be tempting to think that to adopt a moral code is simply to comply with it, i.e. to obey its rules (e.g. Parfit 2011: 405). Today, though, rule consequentialists generally equate adoption with acceptance. To accept a code is to embrace it psychologically, somehow, but someone might embrace rules without following them perfectly because of mistakes, weakness of will, etc.2 In fact, most contemporary rule consequentialists equate adopting a code specifically with ‘internalizing’ it, where this means attaching moral reactive attitudes to the code—including feeling compunction when one contemplates violating its rules, guilt when one does so anyway, and perhaps having some disposition to blame others who break its rules. We might say that for a person to internalize rules is for her conscience to ‘enforce’ them; as Hooker writes, ‘to accept a code of rules is just to have a moral conscience of a certain shape’ (Hooker 2000: 91; see also Brandt 1979: 164–76). There are ‘teaching costs’ involved in instilling or inculcating rules into people, so that they internalize them, and ‘enforcement costs’ associated with punitive reactive attitudes like guilt and blame.3 Rule consequentialists must factor these costs into their calculations.4 If this is what it is for an individual to accept a moral code, then what does it mean for the code to be generally accepted? One obvious answer is that it is for everyone to accept the code.5 Both Brandt and Hooker reject this answer, though, on the grounds that if we cash out general acceptance as unanimous acceptance then the ideal code might not include any rules governing the treatment of ‘dissenters’ who reject it. If everyone accepted some moral code, then there would never be any occasion to act on rules that govern how to treat dissenters. But for a code to include rules that would never be acted on would generally be inefficient; it would increase the code's teaching costs with no offsetting benefits. The ideal code will not be inefficient. III. FIXED-, VARIABLE-, AND OPTIMUM-RATE RULE CONSEQUENTIALISM Rather than admit that the ideal code has nothing to say about how we should treat dissenters, Brandt and Hooker instead opt for an understanding of what it means for a moral code to be generally accepted that does not require unanimity. Hooker's ‘fixed-rate’ rule consequentialism (FRRC) identifies the ideal code with the set of rules whose ‘internalization by the overwhelming majority of everyone everywhere in each new generation’ would be optimal (2000: 32). While acknowledging that the precise figure is somewhat arbitrary, Hooker operationalizes ‘overwhelming majority’ as 90% of the population, so he identifies the ideal code as the one whose acceptance at that fixed level would be optimal (2000: 84). This code would presumably include rules for dealing with dissenters, since 10% of the population is too many people to ignore. Ridge offers two objections to Hooker's FRRC. The first is that it is ‘utopian’, since its rules ‘will be narrowly tailored to deal with the problems generated by non-acceptance of the rules by exactly 10% of the population’, and ‘very different rules’ might be needed if the percentage of dissenters is much higher (Ridge 2006: 245).6 For instance, Ridge hypothesizes, a rule requiring moral proselytizing might be irksome if there are relatively few dissenters but useful if there are many. The second is that FRRC is arbitrary, since once the possibility of formulating the theory in terms of 100% acceptance is abandoned there is no satisfying rationale for why a rule-consequentialist theory should be based on one fractional fixed rate rather than another (Ridge 2006: 246–7). To be clear, Ridge is charging not merely that Hooker's 90% figure is arbitrary but instead that the very idea that the ideal code is the one that it would be optimal for the overwhelming majority to accept is arbitrary. This idea's plausibility, he maintains, depends on presuppositions about moral psychology and the content of morality to which Hooker is not entitled. As an alternative, Ridge proposes a ‘variable-rate’ rule consequentialism (VRRC) according to which an action is right if and only if it would be required by rules which have the following property: when you take the expected [value] of every level of social acceptance between (and including) 0% and 100% for those rules, and compute the average expected [value] for all of those different levels of acceptance, the average for these rules is at least as high as the corresponding average for any alternative set of rules. (2006: 248) Ridge's view amounts to the following: find the function for each code of rules that gives its expected value at every acceptance rate from 0% to 100% (inclusive), then define the ideal code as the one with the greatest area beneath the curve.7 (With identical denominators for every code, averaging is unnecessary.) Against Ridge, Smith objects that while it may not be arbitrary to take account of every possible acceptance rate or level in evaluating competing moral codes, it is unmotivated: ‘It is true that testing a code by its acceptance-[value] at exactly 90 percent acceptance seems arbitrary—but why test its acceptance-[value] at every possible level, when in reality it must be accepted at some level or the other?’ (2010: 416).8 On Smith's alternative formulation, the acceptance-value of each code is still calculated at every acceptance rate, but only their maximum acceptance-values are compared. Smith helpfully names the function that gives the expected value of a moral code's being accepted at every possible rate its ‘acceptance profile’ (2010: 416). When we restrict our attention to the acceptance profiles of ‘plausible’ moral codes, Smith suggests, we will generally see acceptance rates and expected values rising together—but only to a point. ‘[T]he teaching generation can affect what proportion of the population accepts a given code through deft choice of teaching technique’, but reaching very high acceptance rates may require the use of ‘harsher’ techniques (2010: 416). The higher teaching costs associated with these techniques will eventually outweigh the benefits of greater uptake, but this tipping point will occur at a different acceptance rate for each code. Smith's formulation, which she calls ‘optimum-rate’ rule consequentialism (ORRC), selects between moral codes based on these peaks. Whereas VRRC identifies the ideal code as the one whose acceptance profile has the greatest area under the curve, ORRC takes a ‘maximax’ approach. Unfortunately, Smith's ‘official’ statement of the view obscures this. In ORRC, she writes, ‘An ideal code is the code whose optimum acceptance level is no lower than that of any alternative code’ (2010: 418). This definition is misleading, however, even beyond the awkward mix of definite and indefinite articles. An acceptance level, for Smith, is a percentage of the population. This makes the ‘optimum acceptance level’ of a code the percentage at which the acceptance profile reaches its maximum value. Therefore, Smith has framed her definition in terms of the argument at which the acceptance profile function attains its maximum value rather than the maximum value itself. This must be a mistake, since she clearly indicates that codes with lower optimum acceptance levels can be preferable to codes with higher ones (2010: 417–8). I take her intent to be instead to define ORRC’s ideal code as the one whose acceptance profile reaches the greatest maximum value. IV. SMITH’S INDETERMINACY THESIS After offering this alternative to Hooker's and Ridge's formulations of rule consequentialism, however, Smith asserts that all three formulations are fatally flawed. It is impossible to make the comparisons between competing moral codes that they require, she claims, because the expected value of a given percentage of the population's accepting a given code is indeterminate. Call this the ‘indeterminacy thesis’. While Smith is right that all three formulations should be rejected, neither of her two rationales for the indeterminacy thesis ultimately convinces. Smith's first line of argument concerns conditional rules. Ridge argues that VRRC’s ideal code might include rules that are conditioned on the acceptance rate of the code: [P]erhaps the ideal code would include rules for conditions in which ‘virtually everyone accepts the rules’, where ‘most people accept the rules’, in which ‘at least a majority accept the rules’ and where ‘less than half the people accept the rules’ and where ‘virtually nobody accepts the rules’ (2006: 250). Ridge takes this to insulate VRRC against the charge of utopianism, since these rules would help to ensure that its ideal code offers people appropriate moral guidance however widely it is accepted. Smith claims that ORRC’s and (probably) FRRC’s ideal codes would incorporate similar rules (2010: 419–20).9 Smith's demonstration that conditional rules lead to indeterminacy involves an example featuring three industrialists who must each choose between two methods of waste disposal. She shows that there are different combinations of choices, with different consequences, that are compatible with all three industrialists complying with the code of conditional rules. The first rule, for example, is ‘If two other industrialists are discharging waste [into the river], you ought to discharge’. While this is only a toy example with unrealistically specific rules, Smith takes the point to generalize. She concludes that even 100% acceptance of a set of conditional rules need not lead to a determinate outcome (2010: 421–6). Yet the conditional rules in Smith's example differ crucially from Ridge's. Smith's rules refer to the actions of the other industrialists in their antecedents, Ridge's to different acceptance rates of the code. Earlier in her article, prior to arguing for the indeterminacy thesis, Smith mentions two conditional rules in the form of Ridge's that pertain to the industrialist case, e.g. ‘When less than half the people accept the code, dispose of your waste in the river’ (2010: 419). Had she used these rules in her example the argument would not have worked; there is only one outcome compatible with all three industrialists complying with these rules, since who accepts the code—and so which rule's antecedent is satisfied—is settled before anyone decides whether to pollute. Smith therefore gives us no reason to think that that moral rules conditioned specifically on different acceptance rates can give rise to the indeterminacy that she describes. Moreover, it is not clear that either Hooker's or Smith's own version of the ideal code would contain such rules. Take Hooker's formulation, for instance. When we presuppose a 90% acceptance rate, rules with antecedents like ‘less than half the people accept the rules’ will never be followed, so including them in a moral code would be inefficient. Smith's second rationale for the indeterminacy thesis is that we cannot calculate the expected value of any moral code's being accepted at any rate less than 100% unless we know which moral rules the dissenters have internalized, if any, or at least what their motivations are. As Smith observes, the ‘possible ways of rejecting a given code are legion’, and ‘it seems likely that different rejecters will reject any given code in many different ways’. She illustrates this point with a one-rule code, C*, that requires full payment of all debts. A dissenter might instead accept a rule that requires repayment at any percentage higher or lower than 100%, a rule that allows for an exception like ‘partial or delayed repayment of debts when the debtor is hard-pressed for funds’, or a rule that permits ‘killing those to whom one inconveniently owes money’. Or she might reject all rules, ‘in favour of acting purely out of whim or self-interest’ (2010: 427). To embrace any of these options is to reject C*. And C* comprises only one rule. A realistic moral code will contain multiple rules, to reject any one of which in any way is to reject the code. Suppose that we are trying to determine the expected value of a state of affairs in which people generally accept C*. There are numerous reasons that the expected value of this state of affairs might depend on what assumptions we make about the dissenters, but one factor is that the more the dissenters behave like the people who accept C*, the less friction there will be between them (Smith 2010: 430–2). If we are comparing C* with another code C†, then what assumptions we make about the dissenters may determine which code's general acceptance has the highest expected value. So we cannot calculate the expected value of a moral code's being accepted at a given rate (<100%) without knowing what codes the dissenters accept—if any. And we cannot know this if the counterfactual scenario for which we are to do this calculation is characterized only as one in which the code enjoys that acceptance rate, Smith maintains, because in that case there is no fact of the matter about what the dissenters are like. Various distributions of people rejecting the code in different ways will be consistent with this characterization, which fails either to specify explicitly that a particular ‘dissent distribution’ obtains or even to provide us with any basis on which to assign probabilities to different possibilities. The rule consequentialist can always simply stipulate a distribution, but Smith rightly says that there is no ‘reasoned’, non-arbitrary way for them to decide what to stipulate (2010: 428). Since the dissent distribution of the code's being accepted at the rate in question is indeterminate, the expected value of this acceptance rate is indeterminate. There is, however, an answer to Smith's worry, which stems from an insight of her own. Smith observes that the pedagogical choices of a teaching generation will affect the acceptance rate of the moral codes they are trying to instil. But that is not all that these choices will influence. For each code that a teaching generation might instil, there are different combinations of teaching techniques that it might employ. Call a combination of such techniques a ‘system of moral education’. The system of moral education that the teaching generation employs will influence not only the members of the ‘learning generation’ who completely internalize the rules they are being taught but also the dissenters who do not—just as classroom instruction influences not only students who master the material perfectly but also those who only partially absorb it. Experienced classroom instructors have some empirical basis for predictions not only about what percentage of students are likely to completely master the material if it is taught via a certain method but also about what percentages of students are likely to fall short of complete mastery in different ways. Similarly, our collective experience with moral education gives us some basis for predictions not only about what percentage of the population is likely to internalize the rules they are being taught if a particular system of moral education is employed but also what percentages of people are likely failing to internalize them and how, e.g. by internalizing rules that are subtly different in various ways or by failing to internalize any rules whatsoever. If the characterization of the counterfactual scenario within which the expected value of rules’ being accepted includes the fact that a particular system of moral education is in use, therefore, then there is a fact of the matter about what dissent distribution will result. (Or at least there is a probability distribution over different dissent distributions, which is enough for an expected utility calculation.) Admittedly, if the scenario is characterized simply in terms of the code's acceptance rate, this by itself may not uniquely determine the system of moral education in use. Perhaps for some moral codes and some acceptance rates, only one system of moral education could possibly yield that acceptance rate. Sometimes, though, there will be several systems of moral education that would result in a particular code's having the same acceptance rate. However, in such cases rule consequentialists of all stripes might stipulate that the optimal system of moral education that would yield this result is being employed. This stipulation, note, is principled, and Hooker, Ridge, and Smith herself can all make it. Thus, Smith's second argument for her indeterminacy thesis also fails.10 V. CRITIQUING FRRC, VRRC, AND ORRC Unfortunately, we cannot be satisfied with any of these formulations of rule consequentialism. In this section, I will explain why, beginning with ORRC. While I have rejected Smith's general arguments for the indeterminacy of rule consequentialism, her own formulation is indeterminate. I suggested, hopefully charitably, that ORRC’s definition of the ideal code might be rendered ‘The ideal code is the code whose acceptance profile features the greatest maximum value.’ Suppose that we go through an idealized process to find this code. For all potential codes, we consider what states of affairs would result from their being accepted at all different acceptance rates (if they were inculcated via the optimal system of moral education capable of producing that result). For each of these states we determine an expected value. We can assume that one of these states, S*, will have the highest expected value. In S*, let us say, a code C1 has an acceptance rate of R1%. R1% must be C1’s optimum acceptance rate, so C1 has its maximum value in S*. So C1 is ORRC’s ideal code? Not so fast. It is possible that C1 is the only code that anyone accepts in S*, either because its optimum acceptance rate is 100% or because the only people who dissented from it would accept no code whatsoever. Far more likely, though, is that some people who dissent from C1 will instead accept different codes. (Recall here Smith's example concerning the one-rule code regarding the payment of debts.) Assume that in S* R2% of the people accept a code C2. R2% must be C2’s optimum acceptance rate, and hence the maximum value of C2’s acceptance profile is the same as C1’s. So which is the ideal code? Or, if the articles in Smith's definition should be indefinite—‘An ideal code is a code…’—then which ideal code is the authoritative code, the one that determines actions’ moral standings? Nor should we assume that only two codes will be accepted in S*. As Smith documents, there are innumerable ways of failing to internalize a single rule, and any plausible moral code will contain a plurality of rules of at least moderate complexity. Failing to internalize any rule in a code in any way, including internalizing a similar but distinct rule, means failing to internalize the code. It would therefore be no great surprise if in S* a multiplicity of codes was each accepted by only a miniscule percentage of the people. There would likely be strong resemblances between the codes that the great majority of the people accept, but also considerable differences in particulars. While more laborious teaching techniques might produce greater homogeneity, by hypothesis their benefits would not outweigh the costs. This suggests that in the best state of affairs we could produce there would be a multitude of similar but distinct codes that all had the (same) greatest maximum value at their optimum acceptance rates, and hence that by Smith's criterion could all claim to be the ideal code. One way of resolving this indeterminacy might be to identify the ideal code as the one with the highest acceptance rate in S*. This might be supportable if we had reason to expect that in S* there would be one prevalent moral code that had been internalized by a majority of the population, or at least a large plurality. But we do not. C1 might well have the highest optimum acceptance rate at 2.1% or some similar figure, in which case there would be no strong reason to give it precedence over codes whose optimum rates were only slightly lower—especially if a much higher total percentage of people each internalized one of a cluster of codes that differ significantly from C1 but only slightly from each other. This leaves VRRC and FRRC. I concur with Smith's specific objection to Ridge's view. This is perhaps a small point, but it seems especially perverse that on Ridge's formulation the fact that the internalization of a code by literally no one would be especially bad counts against the code (because the variable rate goes all the way down to 0%). However, much force Ridge's objections to Hooker's FRRC have, there is an even larger problem with Hooker's 90% figure.11 In criticizing ORRC, I emphasized how many ways there are for people who are being taught a moral code failing to internalize it. With so many ways deviate, getting 90% of the population to internalize any code perfectly would require investing enormous time and energy into inculcating each of its rules. The teaching costs of this would likely result in FRRC’s ideal code's comprising only a few simple and undemanding rules (and hence looking far less like ordinary morality than Hooker supposes). This objection also has force against Ridge's formulation. VI. UNIFORM-MORAL-EDUCATION RULE CONSEQUENTIALISM The three leading attempts to answer the question of what it means for a moral code to be generally adopted in terms of its acceptance rate all seem to have serious shortcomings. Proponents of this basic approach might continue to search for the ‘right’ acceptance rate (or rates) for rule consequentialists to employ. In contrast, I shall propose an entirely new answer. To introduce it, I will return to the notion of a system of moral education, which I have described loosely as a combination of teaching techniques used by one generation to instil moral rules into another. Of the multitude of possible systems of moral education, some could be interpreted as being intended to instil one specific code of rules into all the members of the learning generation.12 I will call systems of moral education with this property ‘uniform’. I propose to define the ideal code as the code that would be the subject of the optimal uniform system of moral education. I will refer to this as the ‘uniform-moral-education’ formulation of rule consequentialism (UMERC) and to its ideal code as CU. I should say more here about what I mean by a system of moral education; this will be the most programmatic part of my discussion, i.e. the sketchiest. I have described a system of moral education as a combination of teaching techniques. I am presupposing that moral education is a distinct activity, or set of activities, intended to inculcate moral rules. Not all activities that make a difference to what rules people internalize qualify as moral instruction, by this standard, and in practice it may often be difficult to distinguish sharply between what does and does not count. Moral education happens in many places in society: in families, in schools, in churches, in the media (especially those parts aimed at children), and even in university ethics courses. In a uniform system of moral education, all these different agents of moral education coordinate their activities to attempt to instil a single moral code into all the members of the learning generation by the time that they reach adulthood. ‘Non-uniform’ systems aim to instil different codes into different segments of the population or perhaps even into particular individuals. This might involve not trying to instil any rules at all into some people. Employing a uniform system would obviously involve teaching children relatively simple versions of the rules early on and then adding complexity and nuance. Some of CU’s rules, like those of ordinary morality, might be abstract or vague. This would depend on the trade-offs between greater determinacy in the behaviour of the people who internalize those rules and the teaching costs that might be entailed by rules of greater concreteness and specificity.13CU could assign ‘role’ obligations to members of specific groups, but only when it would be optimal to attempt to instil the corresponding rules into everyone. While at first glance this might appear as inefficient as teaching everyone rules that no one will follow, the cases are quite different. If there should be moral rules specific to physicians, for instance, instilling them into everyone will make a difference to what patients feel entitled to from their doctors and how doctors who violate those rules are viewed by the general public. As I noted previously, UMERC involves a new interpretation of what it means for a moral code to be generally adopted. If the teaching generation employs the optimal uniform system of moral education, then there is a very strong possibility that only a small percentage of the learning generation will internalize the code that they are being taught in every detail. Pupils seldom learn precisely what instructors intend to teach, especially when the material is voluminous and complex, and the teaching costs involved in trying to eliminate small discrepancies often exceed the benefits. The optimal uniform system would therefore be likely to produce considerable convergence around CU, in that many people would internalize codes that bear a strong resemblance to it, but with few (if any) perfectly internalizing CU itself. Some variants of CU might even have a higher acceptance rate than CU itself. As should be obvious, one important factor in determining which rules belong to CU will be what sorts of variants on those rules people are likely to internalize when the teaching is not entirely successful. Suppose that some rule would be very desirable for large numbers of people to internalize perfectly, but that it would also be easy for them to internalize instead some corrupted version of the rule whose acceptance would be very undesirable. This rule would be a poor candidate for inclusion in CU. For the teaching generation to employ the optimal uniform system of moral education would therefore not result in CU’s being generally accepted, strictly speaking. Perhaps it would be generally complied with, in the sense that most people's actions would conform to it most of the time, but this would be no truer of it than it would of the many codes that it resembles—and CU certainly would never be universally obeyed. Thus, CU is neither the code whose general acceptance would be optimal nor the code general compliance with which would be optimal. It can, however, be described as the code the general teaching of which would have the best consequences. Hooker, Ridge, and Smith consider teaching costs, but none define the ideal code as the one that it would be best to teach.14 So UMERC represents a fundamental way of reframing rule consequentialism, one that offers a new understanding of what it is for a code to be generally adopted. The notion of general adoption should be capacious enough to include a code's being generally taught. VII. IN SUPPORT OF UMERC My main thrust in this section will be to show that UMERC is not vulnerable to the same objections as the formulations of rule consequentialism that came before it. However, I will first make one affirmative point in its favour, which arises out of Hooker's assertion that FRRC precludes ‘objectionable elitism’: The theory insists there are the same rules for internalization by the overwhelming majority of … everyone, cutting across all social, economic, national, and geographical distinctions. … So members of the educated elite are not asked to mandate different rules for the less well educated. Much less is any deception of the masses endorsed. On the contrary, the rules are to be public. (2000: 85) There is some question about how much importance we should attach to the publicity of moral rules. There is, after all, a Sidgwickian tradition that is very receptive to the possibility of an esoteric elite morality that differs from the morality of the masses (Eggleston 2013; Sidgwick 1981: 489–90). But insofar as publicity is a value, UMERC embodies it further than does FRRC, VRRC, or ORRC. A moral code might be ‘tainted’ by esotericism in two ways. One is through itself being an esoteric code that is instilled only into a select few. The other is through being a ‘mass morality’, the public complement to an esoteric code. For a formulation of rule consequentialism to embody the putative value of publicity completely, it is not enough that its ideal code would not be esoteric. It also must not be the case that this code emerges as the ideal code only because the optimal system of moral education would teach it as a mass morality while teaching different rules to a small elite. While the ideal code of Hooker's FRRC cannot be an esoteric morality itself, given its 90% acceptance rate, the optimal system of moral education capable of achieving that rate might be a non-uniform one that would teach some other moral code to the remaining 10%. They might be the elite. So FRRC is not as antithetical to esoteric morality as Hooker might have us believe. ORRC and VRRC are, if anything, even less so. One or more of the moral codes that meet ORRC’s definition of the ideal code, the codes that are taught in S*, might easily be esoteric. With VRRC the evaluation of each candidate moral code is a much more complicated affair. It requires us, for every possible acceptance rate, to conceive of the optimal system of moral education that would result in that code's being internalized by that percentage of the population. For a single code, therefore, the optimal systems that would deliver some acceptance rates might teach it as a mass morality while the optimal systems that would deliver others might teach it as an esoteric code. In contrast, UMERC’s ideal code is the one that it would be best to teach to everyone. Not only would CU itself be entirely public, but we can be certain that it did not emerge as the ideal code in virtue of how well it complements some esoteric morality. It is not necessarily an objection to these other formulations of rule consequentialism that they are not entirely free of the taint of esotericism. But if the publicity of moral rules is to be valued, then the fact that UMERC more fully embodies this value is a virtue. Hooker or Ridge could always revise his formulation to incorporate the idea of a uniform system of moral education and so claim the same virtue for it. Hooker, for example, might define the ideal code as the one that would be the subject of the optimal uniform system of moral education capable of producing a 90% acceptance rate. But this ‘mashup’ of FRRC and UMERC would still be vulnerable to the objections to FRRC that I have already discussed. The same would be true of a hybrid of VRRC and UMERC. Notice that if Smith were to make a parallel move, then the resulting view would not only have the virtue of fully embodying the value of publicity but would be proof against the indeterminacy objection that I levelled at ORRC. This is because this view would in fact just be UMERC: it would define the ideal code as the subject of the optimal uniform system of moral education, irrespective of the acceptance rate that system produces. I must acknowledge here that Smith does refer at points to the teaching generation's selecting ‘a code’, which suggests that she could implicitly be assuming that they will employ a uniform system of moral education (2010: 417). This may raise the suspicion that UMERC is the view that she really intends ORRC to be, and that I am just putting newish wine in even newer bottles. I cannot entirely discount this possibility. However, while I have reworked Smith's definition of ORRC’s ideal code in order to make it more clearly reflect her evident intentions, I cannot find a way to read it that makes it equivalent to UMERC’s definition of CU. When CU is taught via the optimal system of moral education, people will internalize many other codes besides it; all these codes’ acceptance profiles will have the same maximum value. So to resolve the indeterminacy problem, it is not enough to stipulate that the teaching generation is employing the optimal uniform system. It is necessary to define the ideal code the one this system teaches. As far as I can see, Smith does not do this. (If I have somehow misread Smith, so that she and I are advancing the same view, then I have at least shown that it is not vulnerable to the objections that she considers fatal to it.) Now to the objections. It is a straightforward matter to show that most of the objections that have been raised against the other formulations of rule consequentialism that I have discussed have no force against UMERC. Not everyone in the learning generation will internalize CU, so unlike universal-acceptance versions of rule consequentialism, UMERC will need to incorporate rules for dealing with dissenters. UMERC is not arbitrary, as Ridge alleges that FRRC is, since CU is the best code to teach to 100% of the people in the learning generation. It also avoids FRRC’s requirement of an unrealistically high acceptance rate. Nor is it unmotivated in the way that Smith takes VRRC to be, because it does not select the ideal code based on which code it would be best to teach to every possible percentage of the learning generation. Finally, as I have already shown, UMERC does not share ORRC’s indeterminacy problem.15 This leaves the utopianism objection. The objection, in its most general form, is that the ideal code might offer guidance to people in the learning generation that is ill-suited for the situations that they actually face, because the moral code whose general adoption would be best under certain counterfactual assumptions about codes’ acceptance might be ‘out of step’ with a world in which these assumptions do not hold. At its most extreme, a code's ‘out-of-stepness’ might involve its requiring actions that are literally impossible to perform, perhaps through including rules that are predicated on non-existent institutions and practices. Short of this, the fact that the ideal code is optimal only under counterfactual conditions might lead to its requiring actions whose consequences would be very bad or that our intuitions tell us should not be obligatory, e.g. actions that involve pointless or unfairly burdensome sacrifices. Or it might leave agents free not to do actions that are necessary in order to prevent very bad consequences or that our intuitions tell us should be obligatory. (Maybe there are even milder ways in which a code could be said to be out of step, but the milder they are, the weaker the objection.) FRRC’s alleged out-of-stepness was a product of the fact that the percentage of people who accept the ideal code might be very distant from the fixed acceptance rate that was assumed in determining what its rules would be. Similarly, UMERC might need to contain rules governing interactions with people who have not internalized CU, but since it is formulated on the assumption that everyone will have at least been taught CU, those rules may still be out of step with a world in which that is not the case. In the real world, there is no moral code—let alone CU—that has been taught to everyone. So why should we not expect CU’s rules to be inappropriate for the choices that we face? VIII. IS UMERC UTOPIAN? In this section, I will take at least a few steps towards showing that UMERC is not objectionably utopian. My strategy in answering this objection will be to distinguish alternative possible interpretations of UMERC and to offer some considerations that suggest that the utopianism objection has at most limited force against the best. There are numerous dimensions on which formulations of rule consequentialism can differ, which means that my characterization of UMERC can be filled in in different ways. One dimension concerns the nature of the good. UMERC becomes a rule-utilitarian theory when it is combined with welfarism, for instance. Another dimension concerns what we can call the scope of the population. Hooker, we have already seen, takes there to be one ideal code that is authoritative for everyone everywhere: the one whose general adoption globally would be best. Brandt, in contrast, depicts different societies as having their own ideal codes, each of which is authoritative for it (1988: 350–1). We can call the first view ‘communal universalism’ and the second ‘communal relativism’. I lean towards utilitarianism and a type of communal relativism, but I will not discuss them any further here. I do, however, want to consider another dimension along which ‘rule consequentialisms’ can differ. A rule consequentialist might be a ‘temporal universalist’, identifying the ideal code with the code whose general adoption ‘everywhen’—throughout time—would be best. Or she might be a ‘temporal relativist’. J. S. Mill is an early rule consequentialist who reveals himself to be both a communal and temporal relativist when he says that the liberty principle only binds certain societies, and that it only began to bind them when they attained a certain level of development.16 What I will call ‘open-ended’ temporal relativism says that the ideal code for us is the one whose general adoption from now throughout the future would be best. ‘Closed’ temporal relativism says that it is the code whose adoption for some finite period would be optimal and that it is authoritative only during that period. A temporally universalist version of UMERC (or any other variety of rule consequentialism) would probably be hopelessly utopian. The world that we live in now surely bears no resemblance to the world that we would be living in if the moral code whose general adoption through humanity's past, present, and future would be optimal had in fact been adopted heretofore. That code might presuppose social institutions and practices that we cannot even imagine. A temporally relativist rule consequentialism of the open-ended variety might not require the impossible. Still, since it would need to guide people's behaviours across many generations, during which time institutions and practices will differ widely, it might contain only abstract principles. Maybe this is not a decisive objection; Kantians, for instance, seem content with a moral standard that comprises just one very abstract principle. (Or is it three?) But for rule consequentialists, who are interested in which moral code's adoption would be optimal, it would be a disappointment if the ideal code did not offer fairly determinate guidance in areas in which tasking individual agents with interpreting abstract principles might go badly wrong (e.g. when especially important issues are at stake or where interpersonal coordination is needed). Nor does this open-ended relativism offer, as it might seem to at first glance, an unchanging ideal code. What it would be best to do ‘from now on’ depends on when ‘now’ is: each new generation can ask themselves which code it would be best to adopt starting now, and we should not expect any two of them to reach precisely the same answers. A better interpretation of UMERC is one that incorporates closed temporal relativism, taking the ideal code to be the one whose general adoption would be best over some fairly brief period, e.g. a generation. In this interpretation, UMERC might say that an action is wrong if it would be forbidden by the moral code that would have been taught to the members of the agent's generation by the optimal uniform system of moral education. So, for instance, an action is wrong for me if my parents and the members of their generation should have attempted to instil rules forbidding it into me and the rest of my generation, given the world as it was when I was coming of age. There is little danger that these rules could require the impossible by making obligatory actions whose performance is predicated on non-existent institutions and practices. In this ‘generational’ interpretation, therefore, UMERC will at least not exhibit the most extreme kind of utopianism. Whether this generational version of UMERC might exhibit some less extreme but still objectionable utopianism is too large a question for me to work through thoroughly here. However, the answer will depend in part on how different my contemporaries and I would be had we received this education. The greater the alterations this would have produced, the greater the likelihood that they will be responsible for CU’s containing rules that entail objectionable requirements or permissions—objectionable either because of their consequences or because they are counterintuitive. Certainly not everything about us, not even our motivations, would be changed, since many of the influences that shape people fall outside the scope of deliberate attempts to instil moral rules. Our moral education is only part of what we might call our ‘ethical’ education, which in the comprehensive sense in which I am using the term would include attempts to cultivate different virtues into us, determine with whom we sympathize and how strongly, etc. A generational version of UMERC would determine the rules in CU on the assumption that these other elements of our upbringing remained as they were.17 Not even all the factors that determine what rules we internalize fall within what I am calling moral education, since they are not all intentional efforts to instil rules in us. As I said before, trying to spell out more fully what does and does count not as moral education is an area where much work remains before I can offer an entirely satisfactory statement of UMERC. But in the West, at least, the people of my generation grew up in market economies dominated by large corporations. No doubt capitalist institutions exerted some influence on what moral rules we internalized, e.g. rules around competition and property. But in imagining what it might have been like if our moral education had been different, we should hold these economic institutions and their influences constant. These considerations imply that there are real limits to how different people of my age would be if our elders had deployed the optimal uniform system of moral education. And while I am talking about my generation, the point generalizes; whatever the world is like in 300 years, there is a limit to how different the people alive then would be if they had been on the receiving end of the best uniform system of moral education given the world as it was when they were coming of age. This in turn suggests that UMERC may not be excessively utopian. Indeed, UMERC might even appear to be open to the charge of quietism, of being insufficiently critical of the status quo and ordinary morality. Keep in mind, though, that even the generational version of UMERC takes account of long-term consequences. In thinking about what code it would be best to teach the members of my or any generation, it is necessary to factor in the institutional reforms they would make as a result and the moral rules they would go on to teach their own children (which might not be exactly the same rules that they had been taught themselves). It would then be necessary to factor in what these children would go on to do, and their children, and so on. Mill observes that each generation can make the next only a bit better than it is (1967: 740). But he also maintains the possibility of what John Skorupski labels a ‘virtuous spiral’, in which each generation enacts incremental institutional reforms that allow for additional moral improvements, which lead to additional reforms, and so on (1989: 23). Thinking about what moral education it would be best to deliver to a particular generation means thinking about how to start this spiral turning by launching the population out on the best possible long-range trajectory. Thus, taking it as fixed that the members of my generation would grow up under corporate capitalism, for example, regardless of what sort of moral education we received, does not mean regarding corporate capitalism as a permanent fixture. Admittedly, complying with the rules of CU would be easier if those rules actually had been generally taught. One form that the utopianism objection might take is the complaint that it is unfair to blame someone for failing to follow those rules in cases where the fact that they have not been generally taught makes this especially burdensome and/or requires pointless sacrifice. I assume that the rules of CU would forbid raising most animals for consumption. Had these rules been generally taught, vegan and vegetarian meals would be standard in places like long flights, as opposed to the special requests they are now. Suppose that in the world as it is you requested a vegan meal on a flight, but that due to some factor beyond your control the only meal available is a factory-farmed steak. Unless you eat the steak, you will be hungry and miserable for hours, and uneaten meals will simply be discarded. Under the circumstances, it may seem inappropriate for anyone—yourself included—to blame you for eating the steak. If CU would forbid eating the steak, is this a reason to reject UMERC? Not necessarily. Even if we believe that wrongness should be analysed in terms of the appropriateness of the moral reactive attitudes like guilt, we might say that for an action to be wrong is for these attitudes to be justified unless the agent has an excuse (for example, Darwall 2006: 26–7). And the circumstances of this case might well offer you an adequate excuse. This is speculative, but UMERC would also likely not be objectionably utopian in terms of its guidance for dealing with dissenters. CU would be framed on the assumption that everyone had received the same optimal education. It might therefore permit upbraiding dissenters in ways that would be counterproductive or unfair, since it would not recognize their suboptimal moral socialization as exculpatory, and it might not require as much proselytizing as would be desirable. In contrast with FRRC, though, UMERC recognizes that even a population that has received a uniform education may accept diverse codes. So CU would likely not require a harsh response to most dissenters or prohibit all proselytizing. It might also include abstract principles, e.g. of fairness, that would moderate the treatment of dissenters but that are not specific it. Some societies’ moral education might be radically suboptimal, e.g. teaching that one group is entitled to dominate another. There UMERC might require active opposition to the oppressors, in ways that could be quite dangerous given their numbers. But that morality should require this is not counterintuitive, even if the danger might excuse failures to act. Some, including some rule consequentialists, may reject the temporal relativism of the generational version of UMERC, on the grounds that the same moral rules must be authoritative universally. Hooker approvingly quotes Bernard Gert's assertion that ‘The requirements of morality apply to all rational persons’, a proposition that Gert takes to entail that, ‘moral rules and ideals do not change’ (Gert 1998: 214–6; quoted in 2000: 87–8).18 And even those who do not reject the position outright might regard my move toward closed temporal relativism to be ad hoc, motivated only by the need to escape the utopianism objection. But my approach here coheres well with a metaethics that views morality as a social creation that is ultimately justified instrumentally as a way of addressing different human problems, such as ameliorating interpersonal conflicts. David Wong appeals to such a view of morality, arguing for a sophisticated cultural moral relativism on the grounds that different moral systems might address the same problems satisfactorily (1984). Even someone who rejects Wong's deep cultural relativism and believes that rule consequentialism is the sole correct moral theory might still see this metaethical stance as a reason to favour a version of rule consequentialism that allows for the rules of the ideal code to change over time as the problems that need to be resolved change. Since the problems that we need morality to help us ameliorate largely arise out of fundamental aspects of human nature, we would expect considerable continuity over generations in terms of which moral rules best meet this criterion, but also some incremental change. And rapid alterations in our circumstance, like those we can expect from climate change, might accelerate the pace of change. IX. COUNTEREXAMPLES A reviewer points out that UMERC may have counterintuitive implications for some relatively distant possible worlds. Perhaps a device would produce wonderful consequences if it detects that everyone has been taught the rule ‘Do a jumping jack at noon, no matter what’ but terrible consequences if it detects anyone doing a jumping jack at noon when the actual system of moral education does not teach this rule’.19 UMERC would then entail that doing the jumping jack is obligatory even when this would have terrible consequences. Or, slightly less fantastically, perhaps young people rebelliously internalize the opposite of whatever rules society intends to instil. In this case, UMERC might entail that antisocial behaviour is obligatory for these hellions, since sincerely teaching them rules requiring such behaviour is the only way to prevent it. I might simply bite the bullet and admit that UMERC has counterintuitive implications, perhaps while also denying that we should trust our intuitions about what morality requires in circumstances so different from ours. But a more satisfying reply may be possible, even if here too my remarks will be ‘programmatic’. This reply is that we ought to embrace a relativism that would allow for UMERC to be the appropriate ethical theory ‘for us’ but not for denizens of far-flung possible worlds. This is a more fundamental relativism than communal or temporal relativism about CU within UMERC. The argument for this contention might begin, as the reviewer generously suggests, with the instrumental view of morality mentioned in the last section. If morality is, ultimately, a social construct to help us address problems we face, then why expect the same morality to be warranted for all possible moral communities, regardless of differences in their psychologies and their worlds that would set them radically different problems? Even if we think there are some features that any moral system would have to possess in order to be recognizable as such, and some universal criteria for judging better and worst solutions, ‘local conditions’ could play a major role in determining not only what specific actions were right or wrong or which specific rules agents should adopt but even which theory of morality offers the best account of its requirements. UMERC might emerge as the best account of morality for humanity in our historical epoch, but not for a world peopled by beings with wildly different moral psychologies or containing technologies so far beyond ours as to be indistinguishable from magic. Depending on how like or unlike ours their local conditions are, a different rule consequentialism, a non-rule consequentialism, or even a non-consequentialist theory might be most fitting.20 Saying that all of this might be so is one thing; showing that it is so is another. I cannot do the latter here. However, although I am painfully conscious of leaving an enormous IOU unredeemed, I have at least shown that UMERC represents an advance on the other formulations of rule consequentialism considered here. All versions of the theory must contend with counterexamples that arise out of their implications for distant possible worlds. UMERC is not open to the further objections that threaten FRRC, VRRC, and ORRC.21 Footnotes 1 " More specifically, this is a characterization of ‘ideal-code’ rule consequentialism, with which my discussion will be exclusively concerned. Cf. Miller (2009). 2 " One reason for interpreting adoption in terms of an agent's psychology rather than her behaviour is that this makes it easier for rule consequentialists to answer a perennial objection, namely that rule consequentialism ‘collapses’ into act consequentialism (Miller 2014: 150–2). 3 " Although I will say little about blame, to the extent that the rules that we internalize influence when we blame others our dispositions to blame will be a factor in the expected value of a code's being generally accepted. 4 " Hooker, for instance, argues that because of these costs the ideal code will be less demanding than act consequentialism (2000: 159–74). 5 " See, e.g. Parfit (2011: 377–80). 6 " While Ridge writes specifically about rule utilitarianism, his points generalize to rule consequentialism and I will use the broader term. Sanford Levy also raises this objection, which he calls the ‘Ideal World Objection’ (2014: 598–614). 7 " See also Parfit (2011: 317–9). 8 " As with Ridge, I will substitute ‘rule consequentialism’ for Smith's ‘rule utilitarianism’. 9 " Hooker, it should be noted, denies that FRRC’s ideal code would contain such rules and sees the fact that VRRC’s might as a weakness of the formulation; see Hooker & Fletcher (2008: 351–2). 10 " Shang Long Yeo suggests in effect that even without this stipulation the rule consequentialist might assign probabilities to all of the different systems of moral education that would achieve the acceptance rate in question and calculate an expected utility on this basis (2017). However, there is no non-arbitrary way in which these probabilities could be assigned. As a reviewer notes, the moral education that people receive and the moral rules that they internalize are not the only influences on their behaviour. So we will need to know more about the counterfactual conditions under which codes are being evaluated than acceptance rates, dissent distributions, and systems of moral education to perform expected utility calculations. In some formulations, there may be no facts of the matter regarding this additional information, and they will still be indeterminate. In some, including my own preferred formulation, there will be (see footnote 17). 11 " For a rejoinder to these objections, see Hooker & Fletcher (2008: 351–2). 12 " Or at least those members without significant affective, cognitive, or conative deficits, a qualification that I will leave implicit in what follows. 13 " As Brandt points out, very abstract rules might also entail high teaching costs (1988: 351). 14 " UMERC does resemble a view recently developed by Tim Mulgan, ‘ideal moral outlook utilitarianism’ (IMOU) (2017; see also Mulgan 2020: 16–25 especially). On closer inspection, though, the theories are so distinct as not even to be rivals. Both are framed in terms of what dispositions it would be best for one generation to instil into its successor; Mulgan calls the optimal set of dispositions the ‘ideal moral outlook’. There are, however, several significant differences between the views. IMOU asks what dispositions it would be best for the current generation to instil into the next, not what the current generation should have been taught. The ideal outlook can comprise much more than a code of rules, e.g. virtues and priorities. Whether Mulgan means to require that one uniform outlook must be taught to everyone in the next generation is unclear. And most importantly, at least at this stage of its development IMOU is a pedagogical theory that answers the question of what we should teach, not a moral theory about anyone's obligations (2017: 290). 15 " Additionally, neither the single-rule act-consequentialist code nor a massive code with the same implications could be the best code to teach to everyone, so UMERC does not collapse into act consequentialism. 16 " I defend this reading of Mill in Miller (2010). While Mill does not advocate a view as specific as UMERC, he is deeply concerned with moral education and his rule consequentialism is an important influence on all the formulations discussed here. 17 " This illustrates why UMERC avoids the possible indeterminacy that I mentioned in footnote 10. Its counterfactual scenario is simply actual recent history, albeit with a different system of moral education in place. Note that a rule consequentialist might identify the ideal code with the code that would be taught as part of the optimal ‘uniform ethical education’, but this formulation would be highly vulnerable to the utopian objection. If the teaching generation instilled better character traits, say, then less might be left for the conscience to do. So this formulation's ideal code might be excessively permissive for people as they are. (The same might be true of IMOU were Mulgan to say that the members of the next generation are obligated to obey the rules of the ideal outlook—and even more strongly true were he to say this of the current generation.) 18 " Gert actually intends this as an objection to rule consequentialism, which he takes to be unable to account for the unchanging nature of moral rules. 19 " This is inspired by Podgorski (2018). 20 " This kind of relativism, for instance, might be an implication of a view that is fundamentally contractualist and that entails that what sorts of principles people can reasonably accept or reject depends upon facts about them and their world. 21 " I am grateful for comments from Ben Eggleston, Tim Mulgan and two reviewers for this journal. REFERENCES Brandt R. B. ( 1979 ) A Theory of the Good and the Right . Oxford : Clarendon . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Brandt R. B. ( 1988 ) ‘Fairness to Indirect Optimific Theories in Ethics’ , Ethics , 98 : 341 – 60 . Google Scholar Crossref Search ADS WorldCat Darwall S. ( 2006 ) The Second-Person Standpoint: Morality, Respect, and Accountability . Cambridge MA : Harvard University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Eggleston B. ( 2013 ) ‘Rejecting the Publicity Condition: The Inevitability of Esoteric Morality’ , The Philosophical Quarterly , 63 : 29 – 57 . Google Scholar Crossref Search ADS WorldCat Gert B. ( 1998 ) Morality: Its Nature and Justification . New York: OUP . Hooker B. ( 2000 ) Ideal Code, Real World: A Rule-Consequentialist Theory of Morality . Oxford : OUP . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Hooker B. , Fletcher G. ( 2008 ) ‘ Variable Versus Fixed-Rate Rule-Utilitarianism ’, The Philosophical Quarterly , 58 : 344 – 52 . Google Scholar Crossref Search ADS WorldCat Levy S. ( 2014 ) ‘The Failure of Hooker's Argument for Rule Consequentialism’ , The Journal of Moral Philosophy , 11 : 598 – 614 . Google Scholar Crossref Search ADS WorldCat Mill J. S. ( 1967 ) ‘Chapters on Socialism’ , in Robson J. M. (ed.) Collected Works of John Stuart Mill vol. V , 703 – 53 . Toronto : University of Toronto Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Miller D. E. ( 2010 ) J. S. Mill: Moral, Social and Political Thought . Cambridge : Polity . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Miller D. E. ( 2014 ) ‘ Rule Utilitarianism ’, in Eggleston B., Miller D. E. (eds.) Cambridge Companion to Utilitarianism , 146 – 65 . Cambridge : CUP . Google Scholar Crossref Search ADS Google Scholar Google Preview WorldCat COPAC Miller D. E. ( 2009 ) ‘ Actual Rule Utilitarianism ’, The Journal of Philosophy , 106 : 5 – 28 . Google Scholar Crossref Search ADS WorldCat Mulgan T. ( 2017 ) ‘ How Should Utilitarians Think About the Future? ’, Canadian Journal of Philosophy , 47 : 290 – 312 . Google Scholar Crossref Search ADS WorldCat Mulgan T. ( 2020 ) Utilitarianism . Cambridge : CUP . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Parfit D. ( 2011 ) On What Matters , vol. I . Oxford : OUP . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Podgorski A. ( 2018 ) ‘ Wouldn’t it be Nice? Moral Rules and Distant Worlds’ , Noûs , 52 : 279 – 94 Google Scholar Crossref Search ADS WorldCat Ridge M. ( 2006 ) ‘ Introducing Variable-Rate Rule-Utilitarianism ’, The Philosophical Quarterly , 56 : 242 – 53 . Google Scholar Crossref Search ADS WorldCat Sidgwick H. ( 1981 ) The Methods of Ethics , 7th edn. Indianapolis : Hackett Publishing . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Skorupski J. ( 1989 ) John Stuart Mill . London : Routledge . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Smith H. M. ( 2010 ) ‘ Measuring the Consequences of Rules ’, Utilitas , 22 : 413 – 33 . Google Scholar Crossref Search ADS WorldCat Wong D. ( 1984 ) Moral Relativism . Berkeley : University of California Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Yeo S. L. ( 2017 ) ‘ Measuring the Consequences of Rules: A Reply to Smith ’, Utilitas , 29 : 125 – 31 . Google Scholar Crossref Search ADS WorldCat © The Author(s) 2020. Published by Oxford University Press on behalf of The Scots Philosophical Association and the University of St Andrews. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
TI - Moral Education and Rule Consequentialism
JF - Philosophical Quarterly
DO - 10.1093/pq/pqaa023
DA - 2019-07-01
UR - https://www.deepdyve.com/lp/oxford-university-press/moral-education-and-rule-consequentialism-SPPOvqp3AW
SP - 1
VL - Advance Article
IS - 
DP - DeepDyve
ER -