Obligation, Reason, and Frankfurt Examples
Abstract and Keywords
This article gives an unusual twist to debates about Frankfurt-type examples. It defends the thesis that if agents are to be fit subjects of “morally deontic judgments”, they must have the power to act and to act otherwise. It argues that, if moral praiseworthiness and blameworthiness for actions presuppose that the agents praised or blamed are “fit subjects of morally deontic judgments”, then moral praiseworthiness and blameworthiness would also presuppose the power to act and to act otherwise. In defending these claims, the article makes use of a technical analysis of the notion of moral obligation in terms of accessible possible worlds advanced by Fred Feldman and Michael Zimmerman.
Many believe that Frankfurt examples impugn the principle of alternative possibilities concerning responsibility (PAP-R):
PAP-R: A person is morally responsible for what she has done only if she could have done otherwise.
These examples purport to show that a person can, for instance, be morally praiseworthy for doing something despite not being able to do otherwise, as long as the conditions that render her unable to do otherwise play no role in bringing about her action (Frankfurt 1969). An arresting thesis is that if Frankfurt examples undermine PAP-R, they also undermine the principle that moral obligation requires alternatives:
PAP-O: It is morally obligatory for a person to do something only if she could have done otherwise.
My primary objective in this paper is to resist this thesis. A secondary objective is to motivate the view that even if determinism does not threaten moral responsibility by expunging alternative possibilities, its effacing alternatives may well imperil moral obligation.
I begin with a summary of why the issue of whether there is a requirement of alternative possibilities for the truth of judgments of moral obligation should command the interest even of those convinced by or at least receptive to the view that there is no such requirement for the truth of judgments of moral responsibility.
(p. 289) Why Should We Care about Whether Obligation Requires Alternatives?
There are at least three reasons to be concerned about whether moral obligation requires alternatives. First, many theorists are drawn to the following principles:
Praiseworthiness presupposes Obligation (PO): An agent, s, is morally praiseworthy for doing something, a, only if it is overall morally obligatory or overall morally permissible for s to do a.
Blameworthiness presupposes Wrongness (BO): An agent, s, is morally blameworthy for doing something, a, only if it is overall morally wrong for s to do a.1
Temporarily suspending judgment on whether the thesis that Frankfurt examples impugn PAP-O if these examples impugn PAP-R is true, suppose that there is a requirement of alternative possibilities for moral obligation, right, and wrong (i.e., there is such a requirement for the truth of morally deontic judgments).2 Then if PO and BO are true, praiseworthiness and blameworthiness also presuppose alternatives. This, in turn, galvanizes a traditional worry about the incompatibility of determinism and moral responsibility. Take determinism to be the view that at any instant, there is exactly one physically possible future (van Inwagen 1983, 3). Understand “incompatibilism concerning responsibility” to be the thesis that determinism is incompatible with moral responsibility. Of the many different arguments that have been advanced for this species of incompatibilism, a relatively traditional one is: If determinism is true, then we lack alternatives.3 If we lack alternatives, then we cannot be morally responsible for our behavior. Therefore, if determinism is true, we cannot be morally responsible for our behavior. (The second premise just is a restatement of PAP-R.) PAP-R, in tandem with the assumption that responsibility requires control, and that this control consists of the freedom to do otherwise, provides the bridge between the first premise and the skeptical conclusion. Frankfurt examples, conceived as they originally were as counterexamples against PAP-R, if cogent, tell against this premise. If, however, such examples do not impugn PAP-O, and if there are strong independent reasons in favor of PAP-O, then provided one accepts the principle that praiseworthiness requires moral right or obligation, and one accepts the principle that blameworthiness requires moral wrong, one will not be able to escape the traditional argument for the incompatibility of determinism and responsibility merely by jettisoning the second premise.
The second reason regarding why the issue of whether moral obligation requires alternatives is significant concerns the impact of determinism on obligation. “Incompatibility concerning obligation” is the thesis that determinism is incompatible with moral obligation. If there is a requirement of alternative possibilities for the truth of morally deontic judgments, it would seem that determinism threatens obligation by eliminating alternatives.
(p. 290) Finally, the third reason concerning the importance of whether obligation requires alternatives has to do with the intriguing proposal, defended by a number of people, that although determinism undermines moral responsibility and other moral assessments that are conceptually or logically associated with praiseworthiness or blameworthiness, determinism leaves intact other normative appraisals, particularly appraisals of moral right, wrong, and obligation (Pereboom 2001, 2002, 2007a).4 Again, if Frankfurt examples do not call PAP-O into question, and there are independent reasons to accept this principle, the credentials of this intriguing proposal will have to be reassessed.
Let us proceed by outlining a typical Frankfurt example.
Think of such an example as unfolding in two stages. In Stage 1, an agent, Yasmin, decides to do something, x, and intentionally x-s. We are to assume that whether you are a libertarian or a compatibilist, in your account of free action and moral responsibility, Stage 1 Yasmin is morally responsible for deciding to x (and for x-ing).5 In Stage 2, the scenario is developed in a way in which something ensures that Yasmin (Stage 2 Yasmin) decides to x—this thing supposedly precludes Yasmin from deciding to do other than x—without in any way interfering in Yasmin's deciding to x. We are meant to draw the conclusion that because Stage 1 Yasmin is morally responsible for deciding to x, and because Stage 2 Yasmin does not differ relevantly from Stage 1 Yasmin with respect to deciding to x, Stage 2 Yasmin is also morally responsible for deciding to x even though she could not have refrained from deciding to x. Individual Frankfurt examples may differ by way of what is offered in Stage 2 as the ensuring mechanism. In Frankfurt's original case, a “counterfactual intervener”—Black—who can manipulate the agent's mind, is supposed to turn the trick (Frankfurt 1969, 835–36). Stage 2 Yasmin decides and does exactly what Black wants her to decide and do, and Black never intercedes. Had Stage 2 Yasmin, however, showed any involuntary sign of not deciding to x, Black would have intervened and forced Yasmin to decide to x. Owing to Yasmin's, say, deciding to donate to UNICEF on her own, in the absence of Black's intervention, it seems highly reasonable that Yasmin acts freely and is, let us suppose, morally praiseworthy for deciding to contribute to the charity (and, subsequently, for contributing), despite not having alternative possibilities regarding this decision and action. Thus, it has been thought that Frankfurt examples provide strong prima facie reason to believe that alternative possibilities are not required for moral praiseworthiness or responsibility in general.6
Assume that in Stage 1 Yasmin's deciding to contribute and her contributing are overall morally obligatory for her. If this is so, then it would seem that in Stage 2 Yasmin's deciding to contribute and her contributing should also be overall morally obligatory for her. After all, in Stage 2 she acts in the absence of any intervention (p. 291) from Black in just the way in which she did in Stage 1. So, it may be thought, Frankfurt examples should also undercut PAP-O if they undercut PAP-R.
Elaborating, a theorist who wishes to commission Frankfurt examples to show that PAP-O is problematic may begin by reminding us that obligation is conceptually associated with reasons:
Obligations are tied to reasons (OR): If an agent has a moral obligation to do something, a, then the agent has a (pro tanto) reason to do a (the qualification in the first set of parenthesis in this sentence will be shortly explained).
Imagine that in Stage 1, Yasmin reasons correctly about what she is morally obligated to do. She concludes that she ought morally to contribute to UNICEF. This reason, in conjunction with other causal antecedents of her altruistic action (e.g., pertinent desires) nondeviantly and causally issues in her contributing to UNICEF.
Suppose, now, that in Stage 2, Yasmin reasons just as she did in Stage 1. However, at any moment in her reasoning in Stage 2, she could not have reasoned differently (or so we assume) because of the counterfactual intervener. Crucially, Yasmin still believes that her moral obligation to contribute to UNICEF gives her decisive reason to contribute to this charity. Moreover, on the basis of what she takes to be this reason, she contributes to this charity. Thus, one may conclude that if Frankfurt examples are effective in impugning PAP-R, they are equally effective in impugning PAP-O.
Close attention, however, to the sorts of reason that are conceptually associated with obligation casts serious doubt on the view that Frankfurt examples undermine PAP-O.
Principle OR captures the view that (overall) moral obligation provides us with reasons. I do not know what sort of argument may be advanced to support this principle because it seems no less basic than the principle that obligation requires control or that responsibility requires control. It may simply be that obligations of any kind are conceptually linked to reasons of the given kind (e.g., moral obligations to moral reasons).7 We can, however, say something to clarify the principle. I offer little on the concept of obligation save that in the principle “obligation” refers to all in obligation in contrast to prima facie obligation.
As for the concept of “something's being a reason,” understand “a reason” in the principle to denote an objective pro tanto practical reason. Practical reasons are reasons to have our desires and goals, and to do what might secure these goals. Pro tanto reasons are reasons that can be outweighed by other reasons, as opposed to all-things considered reasons, which cannot be outweighed. If the term “reason” means pro tanto reason, each reason has a certain weight. Suppose that, on a particular occasion, you have several different alternatives. Suppose, moreover, that your pro tanto reasons to act in some way are stronger than your reasons to act in any other way. Then you have most reason to act in this way. We may say that acting in this way is reasons-wise obligatory for you: you, reasons-wise, ought to act in this way. (In more common parlance, we might say that you have decisive reason to act in this way.) Suppose that, on a different occasion, you have sufficient or enough pro tanto reason to act in two or more ways, and no better reason to act in any other way. Then we may say that it is (p. 292) reasons-wise permissible for you to act in either of these ways. Finally, suppose you have most pro tanto reason not to act in a certain way. Then we may say that acting in this way is reasons-wise wrong for you (or that you have decisive reason not to act in this way).8 In what follows, departing from common usage and unless otherwise specified, I take pro tanto reasons to denote objective “agent external” reasons.
Objective pro tanto reasons contrast with subjective pro tanto reasons: An agent has a subjective pro tanto reason to do something if and only if she believes that she has an objective pro tanto reason to do it. To bring out the distinction between objective pro tanto reasons and subjective reasons, imagine that you nonculpably believe that you have most pro tanto reason to take some pills that are in the experimental stage of development because you nonculpably believe, on the advice of the medical team, that the pills will assuage your pain. But, in fact, you reasons-wise ought not to take these pills because (unbeknown to the doctors) taking the pills will kill you. Although you have no objective pro tanto reason to take the pills, you have subjective reason to do so. Pro tanto reasons are also to be differentiated from what may be dubbed “Davidsonian reasons”; these are, roughly, complexes of desires and beliefs.9
Why believe, however, that obligations are tied to objective pro tanto reasons and not either to subjective reasons or to Davidsonian reasons? The crux of the matter is that the view (roughly) that some things are morally wrong or morally obligatory for an agent irrespective of what desires or beliefs that agent has is compelling.10 You may believe, on the evidence available to you, that giving medicine M to a sick patient will cure the patient. But if giving M will in fact kill the patient, you do wrong in giving M. You do wrong despite your subjective reason: You believe (let us assume) that you have a pro tanto reason to give M, whereas in fact you have no such reason. Indeed, you have decisive pro tanto reason not to give M. Likewise, you do wrong despite your pertinent Davidsonian reasons: You desire to cure the patient and you believe that you can cure the patient by administering M; your having of this desire and belief (in conjunction with other germane antecedents of action) causally (and nondeviantly) issues in your giving M. None of this, however, need tell against your act not being wrong for you.
In sum, principle (OR) is to be understood as expressing the view that if you have an objective overall moral obligation to do something, then you have a pro tanto reason to do this thing. This principle, I believe, is true.
A Requirement of Alternative Possibilities for Pro Tanto Reasons
If obligation is tied to pro tanto reasons, then Frankfurt examples against PAP-O are ineffective. The reasoning is: (i) There is a requirement of alternative possibilities for the truth of judgments of O-pro tanto reasons; (ii) Assuming that Frankfurt (p. 293) examples succeed in expunging all pertinent alternatives, Yasmin in Stage 2 could not have had a decisive O-pro tanto reason to contribute to UNICEF; (iii) Given the principle that obligations are tied to O-pro tanto reasons (OR), it is false that in Stage 2 where all her pertinent alternatives have been blocked off, Yasmin-2 had a moral obligation to contribute to UNICEF. I confine attention to (i).
If you have most moral reason to do something, a, and, thus, if morality requires that you do a, then you can do a. In other words, the moral “ought” implies “can.”11 Suppose, now, that you have most (practical) reason to do a; as we said, you ought to do a from the point of view of reason. Then you can do a. You cannot have an “obligation”—it cannot be necessary—from the point of view of reason, for you to do something if you cannot do it.12
So it seems that just as there is an association between the “ought” of morality and “can,” there is a similar association between the “ought” of reason and “can.” Indeed, the moral “ought” implies “can” principle appears just to be a more restricted version of the following general principle:
Reasons-Wise “Ought” Implies “Can” (KR): If one has most reason to do something, a, and, thus, if one reasons-wise ought to do a, then one can do a.
If reasons-wise “ought” implies “can,” I see no reason to deny that reasons-wise “wrong” (and reasons-wise “right”) imply “can” as well. As a preliminary remark regarding what may be provided as support for this view, the moral “ought” implies “can” principle, where “ought” expresses all in moral obligation, can be put in this way:
(Kant's Law): If it is morally obligatory for one to do something, then one can do it; and if it is morally obligatory for one to refrain from doing something, then one can refrain from doing it.
It has been emphasized that moral responsibility requires control; if you are morally praiseworthy or morally blameworthy for an action, you have responsibility-relevant control in performing this action. Likewise, think of Kant's Law as a control principle for moral obligation; if you have a moral obligation to perform an action, you have obligation-relevant control in performing it.
If we conceive of Kant's Law in this way—as a principle of control—then barring persuasive reasons to believe otherwise, there is little reason not to assume, too, that moral “wrong” (and moral “right”) imply “can.” I advance, specifically, three considerations in favor of this view. First, as we have registered, just as moral praiseworthiness and moral blameworthiness require control (or freedom), so does moral obligation, moral wrong, and moral right. The control requirements of blameworthiness, unless we have sound reason to believe the contrary, mirror those of praiseworthiness: both have the same freedom requirements. An essential element of the freedom requirement of these responsibility appraisals is captured by these principles: One is morally praiseworthy for doing something only if one could have done (p. 294) it; likewise, one is morally blameworthy for doing something only if one could have done it. These principles highlight a link between moral responsibility and freedom, a link that I believe holds of conceptual necessity. Likewise, it would seem that the control or freedom requirements of moral obligation, unless we have strong reason to think otherwise, should also be the very ones of moral wrong and moral right. If Kant's Law expresses just one more incarnation of the association between morality and freedom, then, again, in the absence of a special reason to believe otherwise, it should also be the case that the principles in which “wrong” implies “can” and “right” implies “can” express two other instances of this association.
Second, this symmetry in the freedom or control requirements of obligation, wrong, and right is validated by a powerful analysis of the concept of moral obligation that Fred Feldman and Michael Zimmerman advance. The analysis provides a plausible treatment of a wide array of deontic puzzles, sometimes partly in virtue of implying that “wrong” implies “can.” In this account at each time of moral choice, there are several possible worlds accessible to a person at that time: There are, at the time, various ways in which a person might live out her life. For each of these complete “life histories,” there is a possible world—the one that would exist if she were to live out her life in that way. A possible world is accessible to a person at a time if and only if it is still possible, at that time, for the person to see to the occurrence of that world. A world may be accessible to a person at a time, but once the person behaves in some way other than the way in which he behaves in that world, it is no longer accessible—it has been “bypassed.” Once bypassed, a world never again becomes accessible.
Making use of the notion of accessibility, one can say that a state of affairs is possible for a person at a time if and only if it occurs in some world still accessible to the person at that time. Let “Ks,t,p” abbreviate “there is a world accessible to s as of t in which state of affairs p occurs.” “Ks,t,p” is equivalent to “as of t, s can still see to the occurrence of p.”
On this analysis, actions are morally judged not by appeal to the values of their outcomes, but by appeal to the values of the accessible possible worlds in which they are performed. Worlds may be ranked in accordance to a value-relation; each world is as good as, or better than, or worse than, each other world. A world is best if no world is better than it is. For purposes of “value-wise” ranking worlds, one can supply one's favorite axiology. I simply label the relevant value “deontic value.” Some may opt for the view that deontic value consists in intrinsic value; others might resist this view. The analysis can now be stated in this way:
(MO): A person, s, ought, as of t, to see to the occurrence of a state of affairs, p, if and only if p occurs in some world, w, accessible to s at t, and it's not the case that not-p occurs in any accessible world deontically as good as or deontically better than w (Feldman 1986, 37).13
More intuitively (and simplifying somewhat), according to (MO), as of some time, an act is morally obligatory for you if and only if you can do it and (simplifying) it occurs in all the best worlds accessible to you at this time.
(p. 295) (MO) verifies a version of Kant's Law. Allowing K to express the relevant sort of possibility, this implication relation can be stated as follows: MOs,t,p implies Ks,t,p. MOs,t,p means that there is an accessible p-world such that there is no as good accessible not-p world. Hence, there is an accessible p-world. This means that Kstp is true as well. Likewise, given (MO), if as of some time p is wrong for you, p occurs in some world that is accessible to you but (simplifying) not in any of the best worlds accessible to you. So, on (MO), “wrong” implies “can” as well.
Some may press the following objection: Why is it true that “given (MO), if as of some time p is wrong for you, p occurs in some world that is accessible to you but not in any of the best worlds accessible to you?” Why does it follow from what has been said about (MO) that p could not occur in some world that is inaccessible to you? Fair enough. Note, first, that if p is obligatory for you, then (simplifying) p occurs in all the best worlds accessible to you; and that if p is permissible for you, then (simplifying) p occurs in some of the best worlds accessible to you. We may put this point as follows: If (MO) is true, then an essential constituent of the control that both obligation and right require is the ability to see to the occurrence of what is obligatory or right. Again, unless we have special reason to believe otherwise, why should this constituent of “deontic control” differ in the case of wrong? Conversely, one may argue as follows. If something can be morally wrong for you even though you cannot do that thing, why should things be different with right and obligation? As (MO) gives us powerful reason, however, to accept the thesis that “ought” implies “can” and the thesis that “right,” too, implies “can,” one ought not to resist the thesis that “wrong” implies “can.” Second, fulfillment of an obligation or a wrong (I admit that “fulfillment of a wrong” has an odd ring, but allow it to pass) affects how the world will be; fulfilling an obligation makes the world deontically better, whereas fulfilling a wrong makes the world deontically worse. If, by virtue of not being able to perform an action, one cannot affect for better or worse the way the world will be, it seems that that action cannot be obligatory or wrong for one.
Some people may question the thesis that “wrong” implies “can” on the basis of the view that “not wrong” implies “right.” But this is a mistake. On (MO), an act may not be right, or wrong, or obligatory if, for instance, one cannot perform it. Such an act would be “amoral.” So from the fact that an act is not wrong, it does not follow that it is right. Yet others may question the “wrong” implies “can” thesis because they believe that “not wrong” implies “not bad” or maybe even “good.” But this would be another error. It is false that an amoral act or even a right act may be not bad or may not even be good. You may find yourself in an unfortunate situation in which all your options are (deontically) bad but one of them is least (deontically) bad, and because of this, obligatory for you.14
The third reason to endorse the view that “wrong implies “can” has to do with obligations to refrain from doing certain things. Suppose that it is morally wrong for you to do something. Then you morally ought to refrain from doing it. Just as one can be (directly) responsible for something only if one can (or could) intentionally do it, so one can be (directly) obligated regarding doing something only if one can (or could) intentionally do it (see, e.g., M. Zimmerman 2006, 595, 602).15 So (p. 296) if you morally ought to refrain from doing something, then you can intentionally refrain from doing it. Such an instance of an intentional refraining would be an instance of your bringing about an intentional omission.
Intentional omission cannot be accomplished without intentionally doing something (i.e., bringing about a “positive” action) in the place of what one omits to do (see, e.g., M. Zimmerman 1988, 23).16 For instance, if you intentionally refrain from turning on the light, then you bring about something (e.g., you intentionally remain seated) for the purpose of the light's not going on.
Revert to a case in which you morally ought to refrain from doing something. Imagine that you ought not (and, hence, that it is morally wrong for you) to turn on the light. Imagine, moreover, that in your circumstances, it is a necessary prerequisite of your intentionally not turning on the light that you remain seated. Then you ought to remain seated as well. For, if you ought to do something (refrain from turning on the light in the example), and you cannot do this thing without doing something else (remaining seated in the example), because this something else is a necessary prerequisite of your doing the thing that you are obligated to do, then you ought to do this something else as well.17 Now, if you ought to remain seated, then you have obligation-relevant control in remaining seated. We are supposing that an essential element of this control consists in your power or ability to remain seated (Kant's Law). Recall, if it is wrong for you to do something, then you ought not to do it. Because all wrongs are “ought nots,” and “ought nots” cannot be accomplished without bringing about something that is itself obligatory, and, which is, hence, something in the bringing about of which you have obligation-relevant control, all wrongs inherit this very element of control.
To flesh this out, in the example, if you ought not (and, so, if it is wrong for you) to turn on the light, then you ought to remain seated. If you ought to remain seated, you can remain seated. So, your moral obligation that is an intentional refraining (or omission)—an “ought not”—can only be accomplished if you have the power to remain seated. But this “ought not” just is a wrong. You bring about the omission (your not turning on the light) by way—and only by way, as we have assumed—of intentionally remaining seated. There is, in the circumstances, nothing more to bringing about your intentional omission (the “ought not”) than bringing about the relevant “positive” action—your intentionally remaining seated for the purpose of not turning on the light. The control on your part in accomplishing this “ought not” or wrong is exhausted by the control that you exercise in bringing about the “positive” action of your remaining seated (for the purpose of not turning on the light). An essential element of this control consists in your being able to perform this positive action. And this strongly suggests that “wrong” implies “can.”
In sum, we have good reason to believe that each of “moral ought,” “moral right,” and “moral wrong” implies “can.”
Reverting, now, to reasons-wise obligation, the reasons-wise “ought” implies “can” principle, just like Kant's Law, expresses the control that “obligations” of reason require. There is, it seems, no reason to believe that the control requirements of the moral “ought” differ from those of the reasons “ought.” As I previously ventured, (p. 297) Kant's Law is just a special case of the general principle that reasons-wise “ought” implies “can.” And, again, precluding compelling reasons to think otherwise, if reasons-wise “ought” requires a species of control, reasons-wise “right” and reasons-wise “wrong” require this very species of control as well: If it is reasons-wise right (or reasons-wise wrong) for you to do something, then you can do it.
We may now argue as follows. KR states that reason-wise “ought” implies “can.”
Its corollary is
Reasons-Wise “Ought Not” Implies “Can Refrain From” (KRC): If one reasons-wise ought not to do something, a, then one can refrain from doing a.
Moreover, we should accept this principle (Reason-1):
Reasons-Wise “Ought Not” amounts to Reasons-Wise “Wrong” (Reason-1): One reasons-wise ought not to do a if and only if it is reasons-wise wrong for one to do a.
From KRC and Reasons-1 derive:
Reasons-Wise Wrongness Requires Alternatives (Reason-2): If it is reasons-wise wrong for one to do a, then one can refrain from doing a.
Reason-2, in conjunction with the claim that reasons-wise “wrong” implies “can,” establishes that there is a requirement of alternative possibilities for reasons-wise wrongness.
Consider, next, reasons-wise obligation. If it is reasons-wise obligatory for one to refrain from doing something, a, then it is reasons-wise wrong for one to do a (from Reason-1). Moreover, if it is reasons-wise wrong for one to do a, then one can do a (from the reasons-wise “wrong” implies “can” analogue of [KR]: if it is reasons-wise wrong for one to do a, then one can do a.) Therefore, if it is reasons-wise obligatory for one to refrain from doing a, then one can do a. But it is also true that if it is reasons-wise obligatory for one to refrain from doing a, then one can refrain from doing a. In other words, just as there is a requirement of alternative possibilities for reasons-wise wrongness, so there is such a requirement for reasons-wise obligation.
If reasons-wise wrongness and reasons-wise obligation require alternative possibilities, I see little reason to deny that reasons-why rightness requires alternative possibilities. We may conclude that there is a requirement of alternative possibilities for the truth of judgments of objective pro tanto reasons.
“Not so fast!” it may be chided. Regarding “wrong” implies “can,” consider the commandments not to commit adultery and not to covet one's neighbor's wife. Suppose one believes it is morally obligatory not to do these things, that they are morally wrong (as many do). And suppose I cannot commit adultery because I am not married. And I cannot covet my neighbor's wife because she is not in the least (p. 298) bit attractive to me. Most people would say that it remains morally obligatory for me not to do these things, because it is morally obligatory for everyone not to do them, even though some are unable to do them.
I start with a preliminary comment: Suppose you cannot save a child who is drowning because you cannot get to this child in time. Some people may say, or may have the intuition that, still, you ought to save the child. I confess to not seeing any relevant difference between this sort of scenario where it is Kant's Law that is up for assessment, and the scenarios involving adultery or coveting your neighbor's wife in which it is the “‘wrong implies ‘can’ principle” that is the object of evaluation. Should one insist that it is wrong to do the relevant things in the latter scenarios even though one cannot do them, unless one has a compelling story to tell about why the control requirements of wrong differ from those of obligation, one should also insist that it is obligatory for you to save the child in the initial scenario despite your not being able to save the child.
Regarding the latter scenarios it is, minimally, controversial whether, at any time, it is wrong for any person at that time to commit adultery or to covet one's neighbor's wife. Given MO, there does not seem to be any barrier against its being possible that, at some time, I covet my neighbor's wife in some best world accessible to me, and there is no accessible world that is deontically as good as or better than this world in which I refrain from coveting my neighbor's wife. Furthermore, it is one thing to claim that most people would say that when I cannot do these things, it remains morally wrong for me to commit adultery or to covet my neighbor's wife, or that they have the intuition that these things are wrong, but quite another to argue that they are wrong. Perhaps some competitor of MO, in conjunction with the relevant facts (including facts about the substantive account of deontic value being assumed) implies, for instance, that committing adultery is always wrong for any person. But then we should assess whether such a competitor, together with its presuppositions, is superior to MO before we can arrive at an overall verdict about whether this objection is cogent.
Here is another sort of case, this time directed against the “‘ought not’ implies ‘can refrain from’ ” (OCR) principle. William Mann (1983) writes:
If the principle is to stand a chance of being plausible, it must be made compatible with the fact that for many people the present structure of their character renders them unable to avoid doing what they know they ought not to do…. A person whose upbringing involved considerable exposure to bigotry may find that he still sometimes makes judgments about members of other races … which he now knows to be bigoted. He recognizes that he ought not to make such judgments, but his habits are so strong that on some occasions—involving, say, haste and pressure—he cannot avoid making them (80).
Mann's objection to OCR amounts to this: It seems that both the following can be true:
M1:s knows that s ought not to do a.
M2:s cannot refrain from doing a.
(p. 299) But if OCR is true, then it is not possible that M1 and M2 both be true: M1 entails that s ought not to do a but OCR and M2 entail that it is false that s ought not to do a.
The objection, however, is not decisive. For appropriately reformulated versions of M1 and M2 enable us to retain the intuition that M1 and M2 can both be true consistent with OCR's being true. We simply need to pay close attention to the double time indexes of “ought” statements. Suppose that, as of time t1, s can avoid doing a (making bigoted judgments) at a later time t3. Then this can be true consistent with OCR:
M1*:s knows that, as of t1, s ought to avoid doing a at t3.
Now suppose due to haste or pressure:
M2*:As of t2, s cannot refrain from doing a at t3.
Imagine that, as of t2, s's inclination to make bigoted judgments gets the better of s so that as of this time s finds that s cannot avoid making such judgments at t3. Still, M1* and M2* can both be true consistent with OCR's being true.
I want to make a slightly different point, however, about Mann's case. Consider this modification of the case designed to cast doubt on “wrong” implies “can”:
If the principle (“wrong” implies “can”) is to stand a chance of being plausible, it must be made compatible with the fact that for many people the present structure of their character renders them unable to do what they know they ought not to do …. A person whose upbringing involved considerable exposure to the commandments may find that he cannot covet his neighbor's wife.
The objection this time is that M3 and M4 can both be true:
M3:s knows that it is wrong for s to covet s's neighbor's wife.
M4:s cannot covet s's neighbor's wife.
If M3 and M4 are consistent, then it is false that “wrong” implies “can.”
My challenge is simply the following: If one is convinced by this variation of Mann's case, I see little reason for one to resist Mann's original case against OCR.
Or, again, in opposition to the “‘ought implies ‘can refrain from’” principle, one might add that if it is morally obligatory for one to feed a starving child, then one can do so. But I may be of a very compassionate nature, such that when I see the child in such a horrid condition I cannot help but feed it. It remains morally obligatory and hence morally right for me to feed this child. So doing otherwise does not seem to be required for having a moral obligation to do something, even though being able to do it is required. And being able to do something does not seem to be required for having a moral obligation not to do it, even though being able to avoid doing it is required.18
Nevertheless, it is one thing to have the intuition that you morally ought to feed the child even though you cannot refrain; quite another to argue for it. Let us change the (p. 300) case a bit. Suppose, given your psychology, whenever you see a starving child, you cannot but refrain from killing it. Do you do wrong when you kill a starving child even though you could not have refrained? (Note that the question is not whether you bring about something that is deontically bad when you kill the child.) I have my strong doubts that ones does moral wrong in this sort of case, and I have these doubts because moral obligation, just like moral responsibility, requires control. An analogy that comes to mind is this: Some are willing to say or have the intuition that a “willing addict” exercises freedom-level control when shooting up even though her shooting up (on a particular occasion) causally issues from an irresistible desire. I have my misgivings: If the desire is irresistible, on the occasion in question, this addict does not act freely.19 (Whether she is morally responsible for shooting up on that occasion is another matter.)
The claim that “ought” implies “can refrain” is intuitively plausible. M. Zimmerman (2008, 147) proposes that the idea that one can be obligated to do something (refrain from killing the child in the example) that one cannot avoid doing suggests, paradoxically, that morality's demands can be empty or trivial.
Looking to other resources to break the deadlock in the sorts of case given above (the cases involving adultery, coveting one's neighbor's wife, and feeding a starving child) in which intuitions conflict is one way to proceed. What, for instance, do plausible analyses of the concept of obligation imply about whether, for example, “right” implies “can?” There are at least two different analyses and a “deontic system” that independently provide support for this principle. As I have reported, Feldman's (1986) and M. Zimmerman's (1996, 2008) approach both validate this principle as well as Kant's Law. A deontic system models the logical structure of fundamental features of common sense morality. A distinctive feature of Paul McNamara's recently developed deontic system is that it represents among other deontic notions, those of right, wrong, obligation, exceeding the moral minimum (c.f. “action beyond the call of duty” or “supererogation”), and permissible suboptimality (c.f. “suberogation”); as well as responsibility notions such as praiseworthiness and blameworthiness (McNamara 2008 and forthcoming). The model validates Kant's Law and the moral “right” implies “can” principle as well. So we have independent confirming evidence for these principles.
Suppose that being able to do something is an essential constituent of the control that obligation and right require. Then, echoing a point I previously made, I cannot see why a control requirement for wrong should differ in this respect. There is nothing special about wrong. Indeed, the morally deontic concepts of right, wrong, and obligation are interdefinable in this sense: If you take one as primitive, the other two are definable in terms of the primitive. It is curious that we often take “ought” so to speak as “primary.” But this is arbitrary (or so it seems to me and, I presume, some deontic logicians). We could take “wrong” as primary. Then we would say things such as “if it wrong, as of some time, for you to do a, then you can do a” and “if it is wrong, as of some time, for you to refrain from doing a, then you can refrain from doing a.” Deontic logicians should find nothing amiss with proceeding in this way. Again, the primary point is simply that if obligation and right require control, and if an essential element of this control is the ability to do the (p. 301) pertinent thing, then barring persuasive reason to believe otherwise, I do not see why this control element is not also one of wrong.
One may agree that there is a symmetry of sorts between obligation and wrong having to do with control. But that symmetry, it may be pointed out, is captured by (1a) “If one ought to do a, then one can do a,” and (2) “If one ought not to do a, then one can refrain from doing a.” Both “ought” and “ought not” involve an element of control, but the control in the one case is being able to do what you ought to do; in the other case, being able to not do what you ought not to do. This captures both the similarity and the difference between “ought” and “ought not” (i.e., “wrong”). But it does not follow from any of this that (1a*) if one ought to do a, then one can refrain from doing a and (2*) if one ought not to do a, then one can do a.
In response, I concur that because “ought” implies “can,” one should accept the principle that “ought not” implies “can refrain from.” But I have not claimed that it follows from these and only these principles that “ought” implies “can refrain” or that “wrong” implies “can.” Rather, I gave independent reasons to sustain these latter principles. Moreover, regarding obligation, Kant's Law underscores the view that control is relevant to both obligatory commissions and obligatory omissions. This is the symmetry that (1a) and (2) capture. I have argued that right and wrong do not differ from obligation in this very respect: control is required for both permissible or wrong commissions and permissible or wrong omissions. Clearly, (1a) and (2) do not do justice to this symmetry.
To take stock, the following argument encapsulates the essential results (or premises) secured (or affirmed) this far.
(OR) If an agent has a moral obligation to do something, a, then the agent has a pro tanto reason to do a.
(RA) If an agent has a pro tanto reason to do something, a, then the agent could have done other than a.
(C) If an agent has a moral obligation to do something, a, then the agent could have done other than a.
Why Frankfurt Examples Fail to Undermine PAP-O
We can now uncover the principal fault in the reasoning previously summarized for the view that should Frankfurt examples subvert the principle of alternative possibilities concerning responsibility, they should also subvert the principle of alterna (p. 302) tive possibilities concerning obligation. Recall, the pertinent part of the tale involving Yasmin unraveled in this way:
If Yasmin did indeed have a moral obligation to contribute in Stage 1 in which Black is not on the scene, then she ought morally to contribute in Stage 2 as well, and she ought to do so owing to the fact that her reasoning in Stage 2 parallels her reasoning in Stage 1. So she has this obligation in Stage 2 even without having pertinent alternatives.
The major defect in this line of reasoning should now be evident. If Yasmin could not but have contributed to UNICEF in Stage 2, then she could not have had a pro tanto reason to contribute to UNICEF because there is a requirement of alternative possibilities for the having of such reasons. One may allow that she contributes in Stage 2 on the basis of subjective reasons or Davidsonian reasons. But such reasons are not pro tanto reasons, and moral obligation is tied to pro tanto reasons.
In this section, I respond to an objection against the view that Frankfurt examples fail to impugn PAP-O. The objection appeals to these very examples to turn the tables on the principle (OR) that moral obligation is conceptually tied to pro tanto reasons. On this objection, first, one insists that in Stage 2 Yasmin ought morally to donate to UNICEF despite not having alternatives, and this is, again, because she conducts herself no differently in this stage than she does in Stage 1 in which it is assumed that she does have such an obligation. Second, one concedes that obligations are conceptually tied to reasons. Third, one then infers that obligations are not conceptually associated with (objective) pro tanto reasons. The proponent of this objection is, in effect, proposing that careful reflection on Frankfurt examples reveals, or should reveal, that it is more plausible that (1a) Yasmin in Stage 2 is morally obligated to donate to UNICEF than that (1b) obligations are conceptually linked to pro tanto reasons. But if one ventures that Frankfurt examples lend more credibility to (1a) than to (1b), then, it seems, in one's estimation the reasons Frankfurt examples provide to favor the proposal that (1a) Yasmin in Stage 2 ought morally to donate, are stronger than the reasons that (1c) considerations independent of anything having to do with Frankfurt examples provide to favor principles KR, KRC, and Reasons-1. To what factors may we appeal to break this impasse? It would not do simply to dig in one's heels and insist that (1c) is more evident than (1a) or to insist on the reverse. But if we note that the sorts of reason available to Yasmin in Stage 2 of the Frankfurt example are either subjective or Davidsonian (or so grant), then there is reason to be optimistic about the balance of critical judgment being tipped in the direction of (1c). And this reason has to do with a thesis (p. 303) recorded previously: some things are morally wrong or morally obligatory for an agent irrespective of what desires or beliefs that agent has regarding those things.20
Determinism, Obligation, Frankfurt Examples, and Semi-Compatibilism
I conclude with some reflections on determinism's impact on moral obligation. It will be helpful to distinguish between strong and weak alternatives. Suppose agent, s, does action, a, at time, t, in world, w. Agent s has a strong alternative at t if the combination of w's past and w's laws of nature is consistent with s's not a-ing at t. Take it that determinism effaces strong alternatives. Weak alternatives are alternatives that you can have despite determinism's being true. In one conception of weak alternatives, for instance, although Yasmin does one thing—she donates—she would have refrained from donating had she wanted, or tried, or chosen not to donate.
Regarding whether determinism imperils the truth of judgments of moral obligation, we registered that if you have an objective overall moral obligation to do something, then you have a pro tanto reason to do that thing. Moreover, if you have a pro tanto reason to do something, then you could have done otherwise. Thus, if you have an objective overall moral obligation to do something, then you could have done otherwise—you had a pertinent alternative. If moral obligation requires strong alternatives, then owing to determinism's expunging such alternatives, nothing is morally obligatory (or morally right, or morally wrong) for anyone in a deterministic world.
If moral obligation requires only weak alternatives, then determinism will not imperil obligation by effacing the sort of alternative that obligation requires. However, a semi-compatibilist analogue of moral obligation cannot be sustained. To amplify: John Fischer explains that semi-compatibilism (or what we may label “semi-compatibilism regarding responsibility”) is the position that moral responsibility is compatible with determinism even if determinism in incompatible with freedom to do otherwise (Fischer and Ravizza 1998, 52–53; Fischer 2006a, 76–78; 2007a, 56). Fischer and other semi-compatibilists have relied heavily on Frankfurt examples to energize semi-compatibilism. Semi-compatibilists propose that if determinism precludes our ever being able to do other than what we do, the fact that we lack alternative possibilities does not, in itself, threaten moral responsibility. An intriguing element of semi-compatibilism is that it supposedly divorces its account of moral responsibility from any appeal whatsoever to alternative possibilities. Semi-compatibilism, with respect to determinism and moral obligation, states that determinism is compatible with moral obligation quite apart from whether determinism rules out alternative possibilities. This variation of semi-compatibilism, of course, cannot be sustained if moral obligation, as I have argued, is conceptually (p. 304) tied to our having of pro tanto reasons, the having of which presupposes our having access to alternatives. Moreover, semi-compatibilism regarding responsibility cannot be sustained even if one accepts the verdict that Frankfurt examples undermine PAP-R, but one nevertheless accepts the principle that praiseworthiness presupposes moral obligation or moral permissibility (PO), and the principle that blameworthiness presupposes moral wrongness (BO).
In conclusion, I have implicitly proposed that one way to explore whether determinism precludes responsibility, or moral obligation, or yet some other sort of normative appraisal judgments of which are tied to reasons is as follows: first, get clear on the sorts of reason at issue; then inquire into whether the truth of judgments of the pertinent sorts of reason presupposes our having alternatives. This latter inquiry may prove more tractable than some rival strategies, such as directly invoking Frankfurt examples to assess whether, for instance, moral responsibility or moral obligation require alternatives; and it may prove more tractable owing to its turning essentially on assessing principles that command independent appeal.21
(1.) See, for instance, H. Smith (1991, 279); Widerker (1991, 223); Fields (1994, 408–9); Copp (1997; 2003, 286–87); Fischer (2006a, 218). I've argued against these principles. See, e.g., Haji (2002b).
(2.) I use “right” and “permissible” interchangeably.
(9.) Some philosophers have been drawn to the view that no practical reasons are provided by the fact that one desires something. See, e.g., Scanlon (1998); Raz (1999); Parfit (1997, 2001, forthcoming).
(10.) I realize that this claim would be rejected by those people—Bernard Williams (1981a), for example, and more recently Mark Schroeder (2007)—who think that pro tanto reasons in some way depend on desires.
(13.) M. Zimmerman (1996, ch. 2) constructs and defends an analysis similar to Feldman's. In his recent book (2008), he advances a different analysis but one which still validates the “wrong” implies “can” thesis.
(14.) I thank Michael Zimmerman for these observations.
(15.) One has indirect (responsibility-relevant or obligation-relevant) control over something just in case one has control over it by way of having control over something else. One has direct control over something just in case one has control over it that is not indirect. Likewise, one is indirectly responsible for something just in case one is responsible for it by way of being responsible for something else; one is directly responsible for something just in case one is responsible for it but not indirectly so. And one is indirectly obligated regarding something just in case one is obligated regarding it by way of being obligated regarding something else; one is directly obligated regarding something just in case one is obligated regarding it but not indirectly so.
(16.) A more cautious view that, I believe, would suffice for the purposes of this third consideration, is that some intentional omissions cannot be accomplished without intentionally bringing about some “positive” action.
(17.) This principle requires qualification in order to avoid Good-Samaritan-type paradoxes. The following qualification seems adequate: if one cannot do p without doing q (perhaps because q is a logical consequence of p), and if one can refrain from doing q, then if one ought to do p, one ought also to do q. This principle is discussed, among other places, in M. Zimmerman (1996, sec. 2.3) and in Feldman (1990).
(18.) I owe this example, the ones involving adultery and coveting one's neighbor's wife, and the objections with which these examples are associated, to Bob Kane.
(19.) Some may propose that in such a case the willing addict's act of shooting up is indirectly free if, for instance, she freely decided and took appropriate measures to become an addict.
(20.) See note 6 above.
(21.) I thank Michael J. Zimmerman and Bob Kane for their insightful comments on a number of issues that I take up in this piece. This essay was written during my tenure of a 2008–2011 Social Sciences and Humanities Research Council of Canada (SSHRC) grant. I am most grateful to this granting agency for its support.