“Ought”, reasons, and vice: a comment on Judith Jarvis Thomson’s Normativity
- First Online:
- Cite this article as:
- Jay Wallace, R. Philos Stud (2011) 154: 451. doi:10.1007/s11098-011-9738-x
- 142 Views
Judith Jarvis Thomson’s Normativity is a sweeping discussion of normative thought in all of its forms.1 On the interesting and provocative account that she defends, normativity is both ubiquitous and multifarious: evaluative properties are intimately connected with our understanding of the nature of the things to which they are ascribed, and these in turn form the basis for directive thought about how things ought to be.
Thomson’s large thesis in the book is that directives—claims about what people, animals, and objects ought to do—may be reduced to a certain class of evaluatives. On the (decidedly non-consequentialist) theory that she favors, directives are to be reduced to one particular evaluative property, that of being a defective instance of a kind. Omitting various qualifications and refinements, A ought to V, on this theory, just in case A belongs to a kind K that admits of defect, and A would be a defective K if A did not V. The evaluative concept of defect thus undergirds the directive concepts and properties, including not only “ought”, but also “should” and “must” as well (p. 230).
There is much more in Thomson’s book than this reductive account of directives. It includes detailed and challenging analyses of a wide array of normative phenomena, illustrated with examples that are often arresting in their deliberate banality—they force us to see how “rich in normativity” our thinking about even the most ordinary objects tends to be (p. 82). Anyone who grapples with the book will be provoked to re-think a lot of what they have taken for granted about normativity in its various dimensions. In these remarks, however, I shall focus primarily on Thomson’s big picture, her defect-based account of the directives. My main contention will be that Thomson’s account does not capture the basic normative sense of “ought”; I shall sketch an alternative way of thinking about directives of this kind, and explore a number of more specific claims that Thomson makes about normativity through the lens of the contrast between her approach and the alternative account that I favor.
For it to be the case that A ought to V is for it to be the case that there is a directive kind K such that:
(α) A is a K
(β) if a kind K doesn’t V, then it is a defective K
(γ) K is not trumped by any directive super-kind (p. 213)
A directive kind is in turn to be understood as a kind that admits of defect, such that there is a property of being a defective instance of the kind (p. 209). This is the general schema that leads Thomson to say that “it is avoidance of defect that is at the heart of the concept ‘ought’” (p. 211).
The water pressure ought to be strong and constant.
The closet ought to be shifted two feet to the right. (Said on a construction site.)
The meeting ought to begin promptly at 4:10.
The pebbles ought to be removed from the patio.
The toaster ought to be recycled.
These sentences all seem to be perfectly idiomatic, and it is easy to imagine circumstances in which utterance of the sentences would express truths. But it is plain that the truths expressed have nothing to do with the kinds to which the grammatical subjects of the sentences belong.
Thomson’s rationale for ignoring examples of this type, if I understand her position, is that she doesn’t believe them to be genuinely normative claims, in the sense that her theory aims to illuminate. Claims to the effect that “A ought to V” are normative, she suggests, just in case they are about A, and say that A is “called on” or “required” to V (p. 200). She contrasts the normative “ought” in this sense with epistemic “oughts” that might be expressed by sentences of the same grammatical form, whose contents and truth conditions are to be understood very differently (pp. 200–205). If this is meant to be an exhaustive inventory of senses of “ought”, however, it hardly provides grounds for ignoring the kinds of examples I have adduced. On its epistemic reading, “A ought to V”, according to Thomson’s tentative analysis, is true in circumstances in which it is very probable that A will V, and hence in circumstances in which there is good reason for believing that V will happen (p. 204). But the examples I adduced are not of this kind. To say that the toaster ought to be recycled or the pebbles removed from the patio or the closet moved to the right is not to make claims that are epistemic in this way.
It ought to be that the water pressure is strong and constant.
It ought to be that the closet is shifted two feet to the right. (etc.)
The fact that these directive truths admit of formulation in such terms suggests that they do not really involve the normative “ought” after all, and so do not count as potential counterexamples to Thomson’s Directive Thesis.
This suggestion doesn’t appear to help, however. The difficulty is that the directive claims that Thomson paradigmatically takes to be normative seem equally susceptible to expletive formulation (in which the apparent grammatical subject is shifted into the proposition that “ought” operates on). There is no obvious semantic difference between “The toaster ought to toast bread” and “It ought to be that the toaster toasts bread”, and similarly for the examples that involve biological organs and varieties of vegetable. The subject in these examples is a kind of grammatical fiction, just as in the different examples I have adduced, and it is therefore misleading to suggest that the corresponding directive claims are somehow specially about them. Furthermore, the way in which the normative “ought” is supposed to be about its subject seems out of place in the examples from which Thomson begins. Thomson says that the “normative meaning” of “A ought to V” is that “A is called on, or required to V” (pp. 200, 207). But I don’t really know what it is supposed to mean to say that a toaster or a tomato or a pancreas is literally called on or required to do something. It is only creatures capable of thought and agency that we call on to do things, and that may be said to stand under requirements.
Philosophers commonly assume that “ought”, in its basic normative sense, has an explicit or implicit index or argument place for agents, expressing a relation between agents and something else (a proposition, say, or an action or attitude).2 On this agential reading, as I shall call it, directive claims to the effect that “A ought to V” involve A as a logical or semantic and not merely as a grammatical subject.3 Such agential directives are in an interesting sense about the agent who figures in them as their subject, and this is reflected in the fact that they do not appear to be semantically equivalent to expletive reformulations. Thus, “Alfred ought to kiss Betty” does not say the same thing as “It ought to be that Alfred kisses Betty.”4 The latter, after all, is semantically equivalent to “It ought to be that Betty is kissed by Alfred” (which is produced merely through passive transformation of the embedded proposition). And yet we have a firm conviction that “It ought to be that Betty is kissed by Alfred” might be true, without it being the case at all that “Alfred ought to kiss Betty.” Agential directives of this kind seem to be paradigm examples of the normative “ought” that Thomson wants to analyze; their “normative meaning” is to specify something that their subject is genuinely “called on” or “required” to do. I suggest that if we want to understand the special features of the normative “ought”, we need to start with cases that in this way literally implicate the agency of their subjects.
(β′) He/she/they ought to see to it that the toaster toasts bread (the person or group being fixed by the context); or
(β″) The world would be better if the toaster toasts bread (the betterness relation fixed by the context); or
(β′′′) It oughtepistemic to be the case that the toaster toasts bread.5
Suppose that we accept this proposal, as a plausible initial hypothesis about the expletive constructions. It might appear to follow that we should deny that “The toaster ought to toast bread” really is equivalent semantically to “It ought to be that the toaster toasts bread.” Thus, it would remain the case that an object that cannot toast bread is defective as a toaster, even if it inhabits a world in which there is nobody who ought to see to it that it toasts bread, and which would in no respect be better if the object toasted bread. Under these circumstances, Thomson might argue, we would continue to maintain that the toaster ought to toast bread, even while we deny that it ought to be that the toaster toasts bread.
There are a number of things that might be said about this argument; I shall restrict myself to a few brief comments. First, if the proposal under consideration were correct, then we would expect to encounter lots of directives involving objects and artifacts in which expletive formulations do not entail formulations that have the objects and artifacts as their subject (just as we do with directives involving agents). Thus, suppose we accept that “It ought to be that the toaster is recycled” (perhaps because there is some contextually salient agent who ought to make it the case that the toaster is recycled). If the theory we are considering were right, it should not follow from this claim that “The toaster ought to be recycled”, since the latter involves the normative “ought” that is constitutively connected to the notion of defect in a kind. But it seems to me that “The toaster ought to be recycled” is plainly true under these circumstances; that is a perfectly idiomatic way of expressing what the expletive counterpart of the directive claim asserts.
This consideration might not be conclusive on its own, however. Thus, even in the agential case, it could be argued that there is some non-expletive formulation of the directive that is true whenever an expletive equivalent is true. If it ought to be that Betty is kissed by Alfred, then there is a sense in which it is true that Alfred ought to kiss Betty; it is just that there is a different sense in which it would not necessarily follow that Alfred (in particular) ought to kiss Betty. Similarly, Thomson might insist on distinguishing senses in the case involving objects and artifacts and organs. She could say that, while there is a sense in which the toaster ought to be recycled in the example given above, there is a different sense which a claim of this non-expletive form would not be true. The different sense is the one that is distinctively about the toaster; it is the sense in which we talk about what the toaster ought to do qua, or “in virtue of”, being a toaster, as Thomson puts it at one point (p. 210). In this distinctive sense, it is not the case that the toaster ought to be recycled, whereas it is true, by contrast, that the toaster ought to toast bread.
I am willing to grant that Thomson’s paradigm examples can be given a sense that is not equivalent to their expletive counterparts. But I would deny that this is the core normative meaning of “ought”; indeed, once we have isolated them in this way, there are reasons for doubting whether Thomson’s paradigm cases are genuine directives at all. Take “The beefsteak tomato ought to be a big, fat tomato at maturity” (p. 210). Claims of this form serve to identify the functions or purposes or ends of objects of a given kind, saying (as Thomson would agree) what they have to do or to be like if they are not to be defective.6 But we have other natural and idiomatic ways of expressing truths of this sort. We can say, for instance, “The beefsteak tomato is supposed to be a big, fat tomato at maturity”, or “The pancreas is meant to secrete digestive enzymes.” It is not at all clear, however, that these locutions involve directives in any intuitive sense. Granting that the tomato is supposed to be big and fat, and the pancreas is meant to secrete enzymes, it hardly follows that they are “called on” or “required” to do these things or to be these ways. This is partly because it is very hard to make sense of these locutions in application to subjects that are not capable of thought and agency. But it is also because the locutions seem different in sense from expressions that say what something is “supposed” to be like or is “meant” to do; they belong to different semantic categories.
A related point is that “ought” is one of a family of directive expressions; as Thomson herself agrees, it belongs together with “should” and “must”. She proposes that these expressions differ from each other in the “gravity of the defect that is in the offing” if one fails to act on directives of the different kinds (with “should” connecting to milder defects and “must” to defects that are more extreme, pp. 229–230). This would lead us to expect that “must” directives of the paradigm variety should be available in application to non-agential objects and organs and living things, since items of all these kinds are obviously such as to admit of grave defect. But I am struck that “must” directives are completely out of place in these cases, even when the gravest defects are clearly in the offing. Inability to toast bread, for instance, would certainly be a serious defect in a toaster. But if you said to me, “The toaster must toast bread”, I would not know what you are talking about. I conclude that the claims that Thomson takes as her paradigms are not really directive claims at all. They are ways of ascribing functions or purposes or ends to objects of various kinds, saying roughly what they are supposed to do or to be like, insofar as they are instances of those kinds.
To summarize, I have suggested that the basic normative “ought” is the agential “ought”, an expression that takes a being capable of thought and agency as its proper subject, and says about that subject that it is called on or required to do something. Thomson, by contrast, thinks that the basic normative “ought” ranges more widely, applying potentially to any kind of subject that instantiates a directive kind. Let us turn now to the case of “ought” claims that are explicitly about human beings, to see how this difference of approach plays out. Thomson holds that the kind “human being” is clearly one whose instances admit of defect. It follows that the Directive Thesis applies to individuals of this kind, yielding a general schema for understanding directives that are genuinely about those individuals. Thus, if A is a human being, then—very roughly—A ought to do V just in case A would be a defective human being if A did not do V.7
If we accept the agential interpretation of “ought” that I have proposed as an alternative to Thomson’s, things look very different. We cannot simply extend a pattern of analysis that applies to non-agential kinds to arrive at conclusions about what human beings ought to do, because the “ought” that figures in those non-agential applications is not the genuine normative “ought” that we are ultimately interested in understanding. Granting that there is such a thing as being defective as a human being, it is still an open question whether a given human individual ought to do the things that are necessary to avoid this form of defect (in the sense of being “called on” or “required” to do those things).
In support of this way of thinking about the situation, I would note that it leaves space for a substantive question that strikes us as intelligible and even important in deliberation about what to do. Thus individual agents might agree that they would be defective as human beings if they failed to do V, but find that they do not happen to care all that much about this particular form of defect or deficiency. In this familiar frame of mind, they might put the following line of questioning to themselves: “I agree that I’m supposed to do V, insofar as I would be a defective human being if I failed to act in this way. But I don’t yet see that I really ought to do V.” The intelligibility of fundamental normative questions of this kind is precisely what the agential interpretation of “ought” would lead us to expect. According to that interpretation, the normative “ought” is not to be identified with the functional “supposed to” that is connected to defect in a kind. It might well turn out that individuals ought to do what they are supposed to do in this functional sense. If so, however, that will be a substantive discovery in practical reason, not something that is guaranteed to be true by the application of a general schema equally applicable to toasters and beefsteak tomatoes. This substantive issue is precisely what our agents are trying to get clear about when they ask whether they ought to do what they concede would be necessary to avoid being a defective human being.
On Thomson’s interpretation, by contrast, it is very hard to understand what the question even is that our agents might be trying to formulate in this situation. True, Thomson does not officially propose her Directive Thesis as a metaethical claim about the meaning of the normative “ought”. Her account might therefore seem to predict that questions about what one ought to do can have an “open” feel even when one is clear that there is something that is necessary if one is to avoid being defective as a human being. On the other hand, it is also clear that she does not put forward her theory in the spirit of an empirical discovery about the nature of normative properties or states of affairs. Her Directive Thesis is not an account of the contingent truth conditions for “A ought to V”, but an account of “what it is for it to be the case that ‘A ought to V’” (p. 209). Furthermore, the account is defended not on grounds of empirical or explanatory adequacy, but through methods of philosophical investigation that yield “conceptual” discoveries (such as the discovery that “the concept ‘defect’ lies at the heart of … the concept ‘ought’”, p. 230). If her proposal were correct, then people who ask whether they ought to do what they take to be necessary to avoid human defect would seem to be guilty, at the least, of some kind of basic conceptual confusion. But this strikes me as implausible; it does not necessarily seem conceptually confused to put to oneself a question of this form.
The framing of “ought” questions in the context of deliberation about what to do is one of the paradigmatic functions of normative language. This makes perfect sense, on the agential interpretation that I have proposed. The agential “ought” takes creatures capable of thought and action as its proper subject, and deliberation is the process whereby subjects of this kind try to get clear about what they are “called on” or “required” to do, with an eye to adjusting their attitudes accordingly. Similarly, it is only subjects of this kind that we would think to advise about what they should do, where this is the other salient context in which “ought” has a distinctive role to play.8 On the approach Thomson favors, by contrast, there is no special connection of normative language to deliberative and advisory contexts of this kind. Claims that incorporate the normative “ought” can be made about many different kinds of object, and there is no presumptive connection of such claims to actual or possible deliberation on the part of the objects to which they apply. The toaster is obviously not apt to resolve the question whether to toast some bread by reflecting on the question whether it ought to do so; nor do we deliberate on behalf of the toaster about this question (as we sometimes do about animals that are capable of thought and agency, but that cannot frame normative questions for themselves), or offer advice to the toaster about its options. This reinforces my suspicion that Thomson’s normative “ought” is not really a genuine directive at all.9
If A is a human being, then for it to be the case that A ought to Vact is for it to be the case that if A knows at the time what will probably happen if he Vacts and what will probably happen if he does not, then he is a defective human being if he does not (p. 216).
This statement makes the truth of “A ought to do V” depend not on whether A would in fact be a defective human being if A did V, but on whether A would be a defective human being if A did V under a possibly counterfactual condition: that A knows the “objective probabilities” of the various outcomes associated with both V-ing and not V-ing. Thomson contrasts this interpretation of “A ought to do V” with two other possibilities: a “subjective” interpretation, on which the truth of claims of this kind depends on A’s actual beliefs about the actions that are open to A and their outcomes; and an “objective outcome” interpretation, which makes the truth of these claims turn not on what is objectively likely to happen if A Vs or refrains from V-ing, but on what would in fact happen under these conditions.
The general tendency of her approach, if I understand Thomson’s reasoning, is to associate the normative “ought” about human action primarily with contexts of second-person advice and third-person commentary. Against subjective interpretations, she maintains that they yield absurd accounts of what we are doing when we are engaged in normative thought in contexts of this kind (pp. 188–191). If you happen to know that A’s beliefs about the outcome of V-ing are false, then you will be inclined to think that A has reached a mistaken conclusion about what they ought to do, where the error is due in turn to A’s epistemic limitations. If A asked you for advice, you would of course take into account the facts as you know them to be in offering a verdict; you would not suppose that A was inviting you to tell them what they already happen to believe.
Considerations of this kind led Thomson in earlier work to favor the “objective outcome” interpretation of the normative “ought”, which ties truths about whether A ought to do V to conclusions about what would in fact happen if A did V. Thus, if flipping an ordinary light switch would in fact cause a dangerous lightning flash in the house next door, then A ought not to flip the switch, even if A has no reason whatsoever for thinking that this is likely to happen.10 Thomson now thinks this view goes too far in the direction of objectivity, however, suggesting that a person ought to do a thing only if “a human being could have known at the time that the person ought to” do it (p. 198). What matters, on this new account, is not what would in fact happen if A does V, but what it would be possible for a human advisor to anticipate happening if A does V, where this in turn is a question of the objective probabilities.
For what it’s worth, this revision strikes me as an unstable compromise, drawing distinctions that don’t seem to have much genuine significance. Thomson would agree that the truth of “A ought to V” does not depend on there being an actual advisor in the offing who is well-informed at the time about the objective probabilities of the outcomes associated with A’s options for action. We are thus to abstract from the epistemic limitations both of A and of the actual people who might potentially give A advice when we think about the truth conditions of “A ought to do V.” But then it seems arbitrary not to go one step further, abstracting likewise from the epistemic limitations that are built into the finite cognitive situation of merely human observers. In the light switch case, for instance, Thomson would have it that A ought not to flip the switch if there was at the time a significant objective probability of a lightning flash in the house next door.11 This remains the case, even if determining the relevant probabilities would have required detailed investigation by a highly-skilled electrician, using techniques that were perhaps unknown to actual electricians in A’s community (or that would not be developed for another 50 years). Once we have gone this far, though, it seems arbitrary to resist the further degree of abstraction that is involved in moving from objective probabilities to objective outcomes. If it is in fact the case that flipping the switch will cause the lightning flash, then it seems irrelevant that it would have exceeded human cognitive capacities to know that this was likely to occur. We might as well appeal directly to what an omniscient (and therefore non-human) hypothetical advisor would have known about the situation at the time of the action.
My main complaint about both of these objective interpretations, however, is that they ignore the distinctively deliberative role of normative thought. Thomson is right to insist that “ought” is often deployed in giving advice to an agent, or in detached commentary about the agent’s options for action. But on the agential interpretation I would favor, “ought” is also constitutively suited to figure in the agent’s own deliberation about those options. The problem with Thomson’s objective accounts is that they abstract too far from the agent’s deliberative situation, tying the normative “ought” to considerations that may not be so much as available epistemically to the agent whom the “ought” is about. This is not to say that we should adopt a subjective interpretation of “ought” in Thomson’s sense, according to which the truth of “A ought to do V” turns on what A actually happens to believe about the consequences of V-ing. What is relevant, rather, is the evidence that is accessible to A: what there is good reason for someone in A’s general position to believe about the consequences of A’s V-ing.12 This is an objective notion, but it is not as objective as the notion that figures in Thomson’s account; it is connected not to the epistemic situation of an idealized (human or non-human) advisor, but to the situation of the agent who is the logical subject of the “ought” judgment in the first place.
Consider the case in which Alfred has to decide how to deploy a limited number of sandbags to block water that is flowing into two different shafts of a mine.13 Alfred knows that there are ten miners in one of the two shafts, but he doesn’t know whether they are in shaft A or shaft B (nor is there any evidence available to him that would enable him to figure out which shaft they are in). Alfred also knows that he could save all ten miners if he uses the available sandbags to block the water from entering the shaft that they happen to occupy, but that the miners will all die if he uses the sandbags to block off one shaft and the miners turn out to be in the other. Finally, it is known to Alfred that nine miners will definitely be saved if he allows the water to enter both shaft A and shaft B, though the tenth miner (the one who is deepest in whichever shaft they occupy) will die. Under these circumstances, it seems clear that Alfred ought to allow the water into both shafts. Moreover, this remains the case even if there is an actual observer on the scene who knows which of the two shafts the miners occupy—a supervisor above ground, for instance, who has lost radio contact with Alfred down below—and even if Alfred knows that the supervisor has access to better information about the likely outcomes of his actions. Alfred is the agent who is trying to solve a concrete problem about what to do, and it therefore makes sense that the normative “ought” that figures centrally in this context should be tied to Alfred’s epistemic situation in particular.
Thomson would object that this conclusion conflicts with our intuitions about contexts of normative advice.14 If the supervisors were able to re-establish radio contact with Alfred down below, for instance, they would not base their instructions to him on Alfred’s original epistemic situation, but on the better information that is available to them. This is correct, and it is a real challenge for a theory of normativity to account for the use of “ought” in contexts of this kind. But an adequate account must also do justice to the different fact that in the original example, Alfred ought to let the water into both shafts, a consideration that reflects the distinctively deliberative role of normative thought.15
Let me turn, finally, to reasons. In an earlier pass through this general territory, Thomson developed an account of “ought” that left it an open question whether there is most reason for agents to do what they ought to do.16 This was a surprising result; as Thomson acknowledges in Normativity (p. 125–126), many philosophers have taken reasons to be the more fundamental normative notion, so that what makes directives true (when they are true) are facts about agents’ reasons.17 In her new book, Thomson launches a direct attack on this part of the conventional wisdom about the normative. She argues that, far from directive facts turning on facts about reasons, the real relation between directives and reasons exhibits the opposite form of dependence: “facts about reasons for action turn on directive facts” (p. 153). In particular, for X to be a reason for A to do V is for it to the case that X lends weight to the proposition that A ought to do V (p. 153).18 It follows from this analysis that there will always be most reason for people to do what they ought to do—not because directives depend on prior facts about the balance of reasons, but because facts about reasons can trivially be read off of directive conclusions about what people ought to do.19
I want to raise two questions about this large and provocative contention, focusing on its interaction with Thomson’s defect-based account of directives about human action. Omitting qualifications that were discussed above, Thomson holds that the fact that A ought to do V just is the fact that A will be a defective human being if A does not do V. The reasons for A do to V, then, will on her account be considerations that lend weight to the proposition that A will be a defective human being if A does not do V. But it seems to me that there are many reasons for action that are not considerations of this kind. Here are some representative examples. The fact that I have a taste for curried foods is a reason for me to try the new Nepalese restaurant that has opened in my neighborhood. But that fact does not lend weight to the proposition that I would be a defective human being if I failed to try the restaurant.20 Reasons in this domain are apparently independent from considerations that bear on the instantiation of defects in the kind human being. Another set of examples involves moral supererogation. The fact that I would thereby ameliorate human misery and suffering is a reason for me to scale back my consumer activities radically and to donate the entire proceeds to Oxfam. Assuming that this course of action goes well beyond what morality requires, however, the fact that human misery and suffering would be alleviated by it is not a consideration that lends weight to the conclusion that I would be a defective human being if I did not embark on it.21 The point is that there are lots of reasons for action that do not seem to be connected to human defect in the way Thomson’s account evidently requires.
Thomson makes a set of confident claims about what it is to be defective as a human being, interpreting this notion as involving instantiation of the conventional moral vices (216–218). Adapting a phrase from Bernard Williams, we might characterize the resulting account of directives about human action by saying that it attempts to “squeeze the good into the right through the tubes of vice”.22 This approach obviously makes some controversial philosophical assumptions about the nature of moral vice; the idea that the familiar moral vices are all ways of being defective as a human being cannot simply be taken for granted, but requires some elucidation and defense (such as those working in the tradition of virtue ethics have often attempted to provide23). But I want to focus on a different and more specific issue about the vices with my final comment. It seems to me very plausible to suppose that our understanding of many of the ordinary moral vices presupposes an antecedent set of background assumptions about what there is reason for people to do. To take some of Thomson’s own examples (p. 218): the vice of cowardice involves an unwillingness to stand one’s ground in defense of things that there is good reason to care about and to protect; pettiness involves a tendency to attach significance to slights and faults that we have good reason to ignore or overlook; and reckless people put at serious risk things that they have good reason to husband and to preserve (e.g. their own life or that of those they love, or the balances in their pension funds).
If this is right, however, then Thomson’s strategy breaks down. We have reason to do the things that are necessary to avoid cowardice, pettiness, and recklessness, only because it is independently true that there are things that we have reason to stand up for and to overlook and to husband and preserve.24 Furthermore, these independent reasons would appear to ground their own independent directives, directives that cannot be reduced to the evaluative notion of avoidance of vice. There are things that count as cowardice, pettiness, and recklessness, only because it is independently the case that there are things that we ought to stand up for and to overlook and to husband and preserve.
See, for example, Wedgwood (2006). A helpful general discussion, from which I have profited, is Schroeder, “Do Oughts Take Propositions?” (unpublished manuscript).
An agent, in the intended sense, is a creature capable of both thought and action, and so a subject of the kinds of attitudes and actions that admit of assessment in terms of reasons. I return to the subject of reasons below.
Thomson’s defect-based analysis of directives might seem to derive plausibility from its implicit unity, identifying a single underlying meaning that is common to directives about e.g. artifacts and agents. But as this passage confirms, Thomson herself acknowledges several irreducibly different meanings of “ought”: in addition to the defect-based directives, there are also not only epistemic “oughts”, but also “oughts” that are to be understood by reference to thoughts about betterness relations between possible ways the world might be.
Thomson denies that the beefsteak tomato is a “function kind” (2008, p. 210), presumably because being big and fat is not literally a function that tomatoes of this kind could be said to perform (it is a way they can be, not something that they do). Nevertheless, tomatoes of this kind were presumably deliberately bred to have these properties, which are in turn conducive to certain familiar culinary purposes, and this is enough to give application to the language of “defect” and “end” in this context.
Thus Thomson writes: “it is precisely by virtue of what we learn when we attend to directives that are about nonhuman things that we can understand all of the directives, and thus those that are about people as well” (2008, p. 207).
I don’t mean to deny that we entertain directive thoughts in contexts that do not involve either deliberation or advice (in reflection, for instance, about whether Caesar ought to have crossed the Rubicon). The point is just that the agential “ought” is constitutively suited to figure directly in deliberation and advice.
A different position, intermediate between Thomson’s and mine, would take the normative “ought” to be the functional “ought” as it applies to individual living things, so that there are true normative “ought” claims that can be made about tomato plants and gerbils and human beings, but not about toasters and can openers. A view of this kind is suggested by Thompson (2008), part 1.
By contrast with her original “Days End” case, in which the lightning flash is the result of “an extraordinary series of coincidences, unpredictable in advance by anyone” (1990, p. 229); or the Russian Roulette example in Normativity, about which it is said that “even if our world is deterministic, God alone could have known all of the facts about the gun that were going to determine that the barrel would stop spinning when there was no bullet under the firing pin” (p. 198).
I borrow this variant of a familiar case from Niko Kolodny and John MacFarlane, “Ought: Between Objective and Subjective” (unpublished manuscript).
She also thinks that it conflicts with our intuitions about the judgments of better-informed observers who are not in a position to give advice (p. 190). But this seems wrong to me: a third party who knows which shaft the miners are in will still think that Alfred ought to let the water into both shafts.
For a provocative response to the problem of reconciling the “oughts” of deliberation and advice, see Kolodny and MacFarlane, “Ought: Between Objective and Subjective”, a paper that has had a strong influence on my thinking about the normative “ought” and on my argument in the present paper.
A clear expression of this now common view is the following quotation from Joseph Raz: “[t]he normativity of all that is normative consists in the way it is, or provides, or is otherwise related to reasons”, in Raz (1999, p. 67).
The considerations that “lend weight” to propositions, on Thomson’s account (2008, p. 130), are reasons for believing the propositions. Her account could therefore be said to treat reasons for action as reasons for a special kind of belief (namely, a directive belief).
Thus if it is true that A ought to do V, then the considerations X that lend weight to that proposition presumably lend more weight to it than considerations Y lend to the conflicting conclusions they support about what A ought to do. Those considerations X are the reasons why A ought to do V, on Thomson’s analysis; so it follows pretty automatically that there is most reason for A to do V.
Or consider the fact there is a modest itch in the vicinity of my left ear, which seems to be a reason for me to scratch in that general area, even though it lends no weight to the conclusion that I would be a defective human being if I failed so to scratch. Thomson discusses similar examples on pp. 147–148 of Normativity, contending that the considerations at issue lend some weight to weak directives about what an agent “should” do (where this in turn is not to be interpreted as a claim about moral obligation). But this seems implausible if, as Thomson maintains, “should” directives are conceptually linked to being defective as a human being.
Thomson notes that there is plenty of room on her account for other forms of (non-directive) positive evaluation about people who exceed what duty requires of them, evaluations of the kind that might figure in praise (2008, p. 231). But the account doesn’t allow us to treat these evaluative considerations as reasons, since by hypothesis they do not lend weight to the conclusion that the agent who fails to act on them would be defective as a human being.
Williams (1995, p. 190). I should stress that Williams did not use this striking formulation to characterize the general strategy of deriving reasons from facts about the avoidance of vice. He was referring to the different and more specific idea that we might explain the reasons for action of non-virtuous agents by thinking about what the phronimos would do if they were in circumstances of ethical imperfection or vice. Williams found this more specific idea problematic, and his objection to it also locates an interesting question for Thomson, of how she would account for reasons for action that agents have only because they are defective as human beings. But I will not go into this question here.
See, for example, Foot (2001), and (more abstractly) Thompson (2008). I am not sure that Thomson herself would disagree with my claim here; this might conceivably be one of the tasks that she is prepared to concede to “moral theory” on p. 218 of Normativity.
Contrast Normativity, pp. 159–160, where Thomson contends that it is less clear what it comes to for X to be a reason to do something than it is what it comes to for X to lend weight to a directive proposition about human action.
I received very helpful feedback on earlier versions of this paper from Niko Kolodny.