Presently there are two very contentious debates ongoing in normative ethics: the Actualism/Possibilism debate and the Subjectivism/Objectivism debate. I argue that the correct moral theory must be an Objective Possibilist one. In Sections 1 and 2, I explain what is at issue in each of these debates. Then, in Section 3, I argue that Subjectivism implies Actualism; in particular, I argue that the only plausible way of constructing a Subjective Possibilism yields results no Subjectivist could find plausible. And then, finally, in Sections 4 and 5, I argue that Actualism is false; in particular, I argue that Actualism yields implausible verdicts in certain cases involving permissible beneficial sacrifice—cases in which it is permissible to cause harm to some in order to prevent harm from befalling others.
1. Subjectivism and Objectivism
The debate between Subjectivists and Objectivists about morality is a debate about what facts the moral status of an agent’s actions supervenes upon.1 Subjectivists maintain that an act’s moral status supervenes only on the agent’s subjective circumstances.2 3 Objectivists deny this.4 The easiest way to see what is in dispute is to look at a case about which they disagree.
Switch : Chen’s subjective circumstances indicate that flipping the light switch will help Sanchez find her wallet and do nothing else. In fact, unbeknownst to Chen, her objective circumstances are such that if she flips the switch a bomb will go off in Cairo, killing many innocent civilians.Chen has two options:
flip: flip the switch~flip: not flip the switch
According to Subjectivists, it is morally permissible for Chen to flip; everything she believes and all of her evidence indicates that flipping the switch will cause no one harm and will help Sanchez find her wallet. According to Objectivists, it would be morally impermissible for Chen to flip and, furthermore, she is morally obliged to ~flip; flipping the switch will cause a bomb to go off killing many innocent civilians and it is wrong to kill many innocent civilians.
Some try to defuse the debate between Subjectivists and Objectivists by distinguishing two different senses of ‘permissible’. According to this reply, there is both a subjective sense of ‘permissible’ and an objective sense of ‘permissible’. According to the subjective sense of ‘permissible’ it is permissible for Chen to flip in Switch. And according to the objective sense of ‘permissible’ it is not morally permissible for Chen to flip in Switch. Subjectivists and Objectivists are just talking past each other, so goes this reply, because they’re talking about different senses of ‘permissible’.
This attempt to defuse the debate between Subjectivists and Objectivists fails. What is at issue between them is what it is for an action to be permissible in the sense of ‘permissible’ that we are trying to give an account of when doing moral theory. That is, it is the sense of ‘permissible’ over which Utilitarians and Kantians disagree when they offer their respective moral theories. When doing moral theory what we’re after is an account of the notion of permissibility that is of concern to the morally conscientious person when she deliberates about what to do. In her deliberations about what to do, the morally conscientious person is concerned to act permissibly and avoid acting impermissibly. But in her deliberations the morally conscientious person isn’t employing two different senses of ‘permissible’. There is just one sense of ‘permissible’ that the morally conscientious person is interested in when in her deliberations she is concerned to act permissibly and avoid acting impermissibly.5 And so, there is no dissolving of the dispute between Subjectivists and Objectivists by distinguishing senses of ‘permissible’, for it is precisely that particular sense of ‘permissible’ about which they disagree. In other words, the question is: is Subjectivism or Objectivism true of the notion of moral permissibility which is of ultimate concern to the morally conscientious person in her deliberations about what to do? This question cannot be dodged by splitting senses of ‘permissible’.6
So much for the Subjectivism/Objectivism debate.7 Another live debate in contemporary normative ethics is that between Actualists and Possibilists.
2. Actualism and Possibilism
Actualists maintain, and Possibilists deny, that a person’s potential future voluntary behavior—what she will later do were she to do some particular thing now—is relevant to her present moral obligations.8 9 As before, an example will help illustrate what’s in dispute:
Headache : On Monday, Patient has an excruciating headache. Though it will go away on its own in five hours, drug D will cure it immediately. However, as D is a very potent drug, if it is administered on Monday, drug E must be administered on Tuesday in order to counter its side effects, otherwise Patient will die. Doctor can administer D on Monday, but she knows that if she does so, even though she will be able to administer E on Tuesday, because of her own laziness then, she won’t.Doctor has two options on Monday:
give d: administer D to Patient~give d: not administer D to Patient
Possibilists contend that Doctor is morally obliged to give d on Monday. Actualists maintain not only that Doctor is not morally obliged to give d on Monday, but also that she’s in fact morally obliged to ~give d. The Actualist points out that Doctor’s administering D to Patient on Monday would have disastrous consequences—Patient will die on Tuesday if Doctor administers D on Monday. The Possibilist counters that Doctor is morally obliged to do the best she can for Patient and there is a course of action open to her—administer D on Monday and then administer E on Tuesday—in which she cures Patient without killing her.
I’ll say more about what’s at issue in the Actualism/Possibilism debate in Section 4. But for now just note that the Actualist sees an agent’s own potential future voluntary actions as being relevant to what she morally ought to do exactly as she sees those of any other agent.10 The Possibilist, on the other hand, thinks that the way an agent’s own potential future voluntary behavior is relevant to her present moral obligations is very different from the way that of others is. If you will go on to do something horrible to an innocent person, A, if I assist someone else, B, then that provides me with a reason not to assist B; however, if I myself will go on to do something horrible to A, something I would easily be able to avoid doing, if I assist B, that fact gives me no reason not to assist B. The dispute between Actualists and Possibilists, then, is a dispute about how tightly, if at all, morality binds one to one’s own agency across time.
I side with the Possibilist here and will offer arguments against Actualism later. First, however, I aim to show that Subjectivism and Possibilism make uncomfortable bedfellows.
3. The Untenability of Subjective Possibilism
Actualism and Possibilism are theses about the relevance of certain objective facts to a person’s moral obligations—what she will do, in the case of Actualism, and what she will be able to do, in the case of Possibilism. Subjectivism is a thesis according to which all that is relevant to what a person morally ought to do is what she believes about, or what her evidence indicates about, the world,11 and not any objective facts. Actualism and Possibilism are thus both, strictly speaking, incompatible with Subjectivism. You might think, however, that there are Subjective analogues of Actualism and Possibilism and that there is as live a debate between them as there is between Objective Actualists and Objective Possibilists. This is not the case. The only way I can see of translating Possibilism into a Subjective moral framework results in a theory that is untenable by any Subjectivist’s lights. Thus, if Subjectivism about moral obligation is true, then Subjective Actualism is the only game in town.
How exactly should Actualism and Possibilism be understood from a Subjectivist perspective? Start with Actualism. For Objective Actualism, one’s own potential future voluntary actions are relevant to one’s moral obligations in the very same way as are any other possible future events. So too, then, should this be the case for Subjective Actualism. According to the most plausible version of Subjective Actualism, however, possible future events are relevant to one’s own moral obligations by entering into one’s expected deontic value calculations in the standard way: what an agent morally ought to do is choose the alternative available to her at a time, t, such that taking it has the highest expected deontic value of all the alternatives available to her at that time.12 And the expected deontic value of each of her alternatives at t is the sum of the products for each of the possible total outcomes of that alternative of the probability of the outcome’s occurrence given that the alternative is chosen and the deontic value of the outcome’s occurrence.13 For a Subjective Actualist, then, the future voluntary actions of an agent will enter into this calculation in just the way the occurrence of any other event will.
As an example, consider Figure 1, which illustrates the tree of possible futures for some choice of an agent, S, at time, t 0 . The node, a, represents S’s choice between two options, a 1 and a 2 at t 0 . There are five other nodes on the tree, two of which—c and d—correspond to potential future choices of S; the other three remaining nodes—b, e, and f—correspond to potential non-choice branching events. The deontic values of the various possible outcomes, x 1 – x 9 , represent the cardinal rankings of the various outcomes according to the moral theory, whichever it so happens to be, that S’s evidence supports.14 Associated with each event is the probability of its occurrence conditional on the occurrence of all the events preceding it along the tree. So, for example, the probability associated with the segment, b 1 , connecting nodes b and d, P(b 1 ⏐a 1 ), is the probability of the occurrence of b 1 at t 1 conditional on the occurrence of a 1 at t 0 . And the probability associated with the segment, f 2 , extending out of node f toward the end of the tree, P(f 2 ⏐a 2 &c 2 ), is the probability of the occurrence of f 2 at t 3 conditional on the occurrence of both a 2 at t 0 and c 2 at t 2 . The expected deontic value for each of S’s two choices, a 1 and a 2 , can thus be straightforwardly calculated:
| EDV(a 1 ) | = | P(b 1 ⏐a 1 )P(d 1 ⏐a 1 &b 1 )(x 1 ) + P(b 1 ⏐a 1 )P(d 2 ⏐a 1 &b 1 )(x 2 ) + P(b 1 ⏐a 1 )P(d 3 ⏐a 1 &b 1 )(x 3 ) + P(b 2 ⏐a 1 )(x 4 ) |
| EDV(a 2 ) | = | P(c 1 ⏐a 2 )P(e 1 ⏐a 2 &c 1 )(x 5 ) + P(c 1 ⏐a 2 )P(e 2 ⏐a 2 &c 1 )(x 6 ) + P(c 1 ⏐a 2 )P(e 3 ⏐a 2 &c 1 )(x 7 ) + P(c 2 ⏐a 2 )P(f 1 ⏐a 2 &c 2 )(x 8 ) + P(c 2 ⏐a 2 )P(f 2 ⏐a 2 &c 2 )(x 9 ) |
And according to the Subjective Actualist, whichever of a 1 and a 2 S ought to take will be determined by which of EDV(a 1 ) and EDV(a 2 ) is higher. (If EDV(a 1 ) = EDV(a 2 ) then both a 1 and a 2 are permissible.)
How about Subjective Possibilism? Objective Possibilism treats the moral relevance of all future events, except the future actions of the agent herself, in exactly the same way as Objective Actualism does. Likewise, Subjective Possibilism should adopt the same moral framework as Subjective Actualism with respect to all possible future events except the possible future voluntary actions of the agent herself. Whereas the Subjective Actualist thinks that the likelihood that the agent herself will perform some future voluntary action, at t 2 , given that she makes a particular choice, a 1 , among her alternatives at t 1 is relevant to whether she morally ought to choose a 1 , the Subjective Possibilist thinks it isn’t. All that matters for the Subjective Possibilist is what she could do at t 2 given that she chooses a 1 at t 1 . (Or, more precisely, all that matters for the Subjective Possibilist is what she believes at t 1 , or what her evidence at t 1 indicates, that she could do at t 2 given that she chooses a 1 .) In particular, all that matters for the Subjective Possibilist is which of the alternatives that might be available to her at t 2 were she to choose a 1 at t 1 is such that it would have the highest expected deontic value were she thereafter at all times to always do what has the highest expected deontic value at those times.
Here, then, is how to think of how the Subjective Possibilist’s moral theory differs from that of the Subjective Actualist’s. For an agent, S, at a time, t, there is the corresponding tree of possible futures from which the Subjective Actualist determines the expected deontic value of each of S’s alternatives at t. The Subjective Possibilist takes that tree and “prunes” it back in the following way: at every node in the tree corresponding to a possible future choice of S’s, the Subjective Possibilist chops off every branch except the one corresponding to the choice with the highest expected deontic value at that time assuming she always chooses the option with the highest deontic value thereafter. (In effect, this chopping off procedure is equivalent to setting the probability of the selected branch—the one not “chopped off”—to 1 and setting the probabilities of all the other branches—those “chopped off” —to 0.) And whereas the Subjective Actualist holds that the alternative which the agent ought to take is the one with the highest expected deontic value, the Subjective Possibilist, on the other hand, holds that the alternative which she ought to take is the one with the highest expected deontic value*, where the expected deontic value* of an alternative is calculated in exactly the same way its expected deontic value is except that the probabilities used are those given by the pruned tree of possible futures. This, it seems to me, is the most natural way of understanding Subjective Possibilism within the most plausible Subjectivist moral framework extant, namely, the expected deontic value framework.
Return to the example illustrated in Figure 1. If we assume that the option at node d with the highest expected deontic value as of then is d 3 and the option at node c with the highest expected deontic value as of then is c 2 , then the resultant pruned tree of possible futures is illustrated in Figure 2. (The dashed lines in Figure 2 represent the branches of the original tree which have been “chopped off” in the pruning process.) From this pruned tree we can calculate the expected deontic value* for each of S’s two choices, a 1 and a 2 :
| EDV*(a 1) | = | P(b 1⏐a 1)(x 3) + P(b 2⏐a 1)(x 4) |
| EDV*(a 2) | = | P(f 1⏐a 2&c 2)(x 8) + P(f 2⏐a 2&c 2)(x 9) |
And according to the Subjective Possibilist, whichever of a 1 and a 2 S ought to take will be determined by which of EDV*(a 1 ) and EDV*(a 2 ) is higher. (If EDV*(a 1 ) = EDV*(a 2 ) then both a 1 and a 2 are permissible.)
Though this is how Possibilism is best understood from within the most plausible Subjectivist moral framework, even so understood, Subjective Possibilism is highly implausible. To see why, consider the following case:
Liverwurst : A, B, and C are drowning. At t 1 , Ruiz can save A or do nothing. If Ruiz saves A at t 1 , she will fall asleep between t 1 and t 2 , and so though A will live, B and C will both die. If Ruiz does nothing at t 1 , then, at t 2 , she will face a choice of ordering a liverwurst sandwich or not. At t 1 , Ruiz knows the following five facts:
(1) if she saves A at t 1 , B and C will die,
(2) if after doing nothing at t 1 she chooses to order a liverwurst sandwich at t 2 , some third party will save A, B, and C,
(3) if after doing nothing at t 1 she chooses to not order a liverwurst sandwich at t 2 , A, B, and C will all die,
(4) if she chooses to do nothing at t 1 , between t 1 and t 2 she will forget both that A, B, and C are drowning and that choosing to order a liverwurst sandwich at t 2 will result in their being saved from drowning, and
(5) if she chooses to do nothing at t 1 , her hatred of liverwurst will persist from t 1 to t 2 , and so, though she will be able to choose to order a liverwurst sandwich at t 2 , the probability that she will then do so is 0.00000001.
According to Subjective Possibilism, what Ruiz morally ought to do at t 1 in Liverwurst is to do nothing. This is because the pruned tree of possible futures for her choice at t 1 consists of two non-forking branches on the first of which she saves A at t 1 but B and C both die, and on the second of which she does nothing at t 1 and orders a liverwurst sandwich at t 2 thereby unwittingly causing the saving A, B, and C. (The pruned tree of possible futures for Ruiz’s choice at t 1 in Liverwurst is illustrated in Figure 3.) And this is because on the pruned tree of possible futures the branch where she fails to order the liverwurst sandwich at t 2 after doing nothing at t 1 is chopped off, for not ordering the liverwurst sandwich at t 2 after doing nothing at t 1 has a lower expected deontic value than does ordering it. So, according to Subjective Possibilism, choosing to do nothing at t 1 has the highest expected deontic value*, and so that is what Ruiz morally ought to do at t 1 . I submit, however, that no one with Subjectivist sympathies will accept that what Ruiz morally ought to do at t 1 in Liverwurst is do nothing; anyone with Subjectivist sympathies will maintain that what Ruiz morally ought to do at t 1 , given everything she knows then, is to save A.
Liverwurst highlights Subjective Possibilism’s troubles with cases in which one’s evidence changes over time. It can’t deliver plausible verdicts about what an agent ought to do in such cases. And this is especially so when the agent knows in advance that her beliefs and evidence about the outcomes of her actions will be defective in the future. That Ruiz ought to do nothing at t 1 given everything she knows at that time in Liverwurst just seems untenable from the Subjectivist perspective. What’s more, as Ruiz’s future choice between ordering a liverwurst sandwich or not won’t, according to the Subjectivist, be an occasion for her to comply with or flout the demands of morality—at the time of choice, from her perspective it will be a perfectly benign choice—the probabilities regarding how she would in fact choose, it seems, should matter, from any plausible Subjectivist perspective. Because behind Possibilism lies the thought that in terms of one’s future actions one should behave in accord with the dictates of morality then, and cases like Liverwurst are ones in which one’s future action doesn’t stand a chance of flouting the dictates of morality from one’s perspective then, given one’s evidence of what one’s evidence in the future will be, the evidence of how she will act in the future in such cases, it seems, should not be screened off.15
Now, it might seem that this problem for Subjective Possibilism is easily avoided. The problem arises because the pruning procedure as applied to Ruiz’s choice at t 2 had it that ordering the liverwurst sandwich at t 2 has the highest expected deontic value* from then on. But that’s true only in virtue of the evidence Ruiz has as of t 1 about what would happen were she to make whatever choice she might make at t 2 . After Ruiz has forgotten about the three drowning individuals at t 2 , however, Ruiz’s evidence then, at t 2 , indicates that ordering the liverwurst sandwich won’t have the highest expected deontic value. Perhaps, it might be suggested, the Subjective Possibilist’s pruning procedure should not be based on the evidence of the agent at the time of the choice with which the tree of possible futures is associated, namely, t 1 , concerning the results of each of her possible choices at each point of pruning, but, rather, it should be based on the agent’s evidence, at t 1 , concerning what at each point of pruning her evidence then will be concerning the results of each of her various alternatives then. Maybe so modifying the pruning procedure will result in a version of Subjective Possibilism that evades the Liverwurst-based objection.
This proposed modification of the Subjective Possibilist pruning procedure won’t do, however. Consider:
Insult : A, B, and C are drowning. At t 1 , Nguyen can either sing a song or do nothing. If Nguyen sings a song at t 1 , she will fall asleep between t 1 and t 2 , and so A, B, and C will all die. If she does nothing at t 1 , then, at t 2 , she’ll face a choice whether to insult Mbeki’s shoes, thereby causing Mbeki some mild distress, or not. At t 1 , Nguyen knows the following five facts:
(1) if she sings a song at t 1 , A, B, and C will all die,
(2) if, after doing nothing at t 1 , she does insult Mbeki’s shoes at t 2 , some third party will rescue all three of A, B, and C,
(3) if, after doing nothing at t 1 , she does not insult Mbeki’s shoes at t 2 , then A, B, and C will all drown,
(4) if she does nothing at t 1 , between t 1 and t 2 she will forget both that A, B, and C are drowning and that choosing to insult Mbeki’s shoes at t 2 will result in their being saved from drowning, and
(5) if she does nothing at t 1 , her hatred of Mbeki and her intense desire to insult Mbeki’s shoes will persist from t 1 to t 2 , and so, though she will be able to refrain from insulting Mbeki’s shoes and causing Mbeki some mild distress at t 2 , the probability that she will then insult Mbeki’s shoes is 0.9999999.
According to the proposed revision of Subjective Possibilism, it is permissible for Nguyen to sing a song at t 1 . It will have this result because when pruning the tree of possible futures associated with Nguyen’s choice at t 1 , at t 2 , the expected deontic value* calculation as of t 2 is carried out not with respect to Nguyen’s evidence at t 1 about what will happen were she to make whichever choice she makes at t 2 , but, rather, with respect to Nguyen’s evidence at t 1 about what her evidence at t 2 will be about what will happen were she to make whichever choice she makes at t 2 . And with respect to that evidence, Nguyen’s choosing not to insult Mbeki’s shoes at t 2 will maximize expected deontic value* as of t 2 ; the evidence Nguyen knows at t 1 she will have at t 2 will indicate that insulting Mbeki’s shoes will cause Mbeki some distress and benefit no one. But if that’s the case, then the resulting pruned tree of possible futures for Nguyen’s choice at t 1 will consist of the branch in which Nguyen sings a song at t 1 and the branch in which she does nothing at t 1 and does not insult Mbeki’s shoes at t 2 , both of which have the same expected deontic value*—both involve A, B, and C dying. (The pruned tree of possible futures for Nguyen’s choice at t 1 in Insult is illustrated in Figure 4.) And so, according to this revised version of Subjective Possibilism it is morally permissible for Nguyen to sing a song at t 1 . I take it, however, that no Subjectivism worth its salt will have that consequence; anyone with Subjectivist sympathies will hold that what Nguyen morally ought to do at t 1 , given everything she knows then, is do nothing.16
As I see no other plausible way of understanding Subjective Possibilism, and these ways of understanding it yield wildly implausible consequences by Subjectivism’s own lights, if Subjectivism about moral obligation is true, then Subjective Actualism, and not Subjective Possibilism, must be correct.17 18
If all of this is right, then Subjectivism implies Actualism. In what follows, I argue that Actualism is false. If I establish this, I will thereby have established that the correct moral theory must be an Objective Possibilist one. (Unless otherwise indicated, in my argument against Actualism below I consider only situations in which the agents in question have full knowledge. So, I do not need to concern myself with the distinction between Objective Actualism and Subjective Actualism, for in cases of full knowledge the two theories are extensionally equivalent.)
4. The Standard Challenge to Actualism and Actualist Replies
The main Possibilist complaint against Actualism is straightforward: were Actualism true, one could get out of a present moral obligation simply in virtue of one’s own potential future voluntary, and entirely avoidable, moral wrongdoing; “but that,” so says the Possibilist, “just isn’t how morality works”.19 For example, if Actualism is true, Doctor can get out of having a moral obligation to give Patient drug D on Monday in Headache just because were she to do so, she’d later horribly wrongly allow Patient to die. But one can’t just get off the moral hook in that way.
To this charge the Actualist has the following ready reply: yes, if Actualism is true, then one can indeed get out of having a moral obligation in virtue of one’s own potential future moral failings, but that’s only when and because those potential future moral failings would be so morally disastrous. What morality cannot allow is that one be morally required to act in some way at a time, t, that will lead to one’s acting even more morally wrongly overall from t onward than one would act by not doing what morality requires one to do at t.20 The Possibilist fails to fully appreciate, says the Actualist, that allowing Doctor to give d on Monday in Headache would be to allow her to do something that will lead to her acting horribly morally wrongly on Tuesday. And, what’s more, letting Patient die on Tuesday after giving her D on Monday is far more seriously wrong than not relieving Patient of the five hours of excruciating pain on Monday were that what morality required. If morality could require people to do that which would lead them to act even more morally wrongly overall than not doing what it requires, morality would be seriously perverse. And morality is not perverse.
Furthermore, the Actualist might point out, no morally conscientious person would, were she in Doctor’s shoes on Monday in Headache, give d on Monday.21 Because one could indeed be a morally conscientious person on Monday and ~give d then in Headache, then it must at least be permissible for Doctor to ~give d in Headache, for no morally conscientious person would ever knowingly do what was morally wrong—moral conscientiousness precludes ever knowingly doing what is morally wrong.22 And because no person, were she being morally conscientious on Monday, given her knowledge of her own future potential morally unconscientious behavior, would give d on Monday in Headache, then it must be that to give d on Monday in Headache would be morally wrong, for if it weren’t it would have to be possible for someone who was being morally conscientious then to do so.
The Possibilist attack upon Actualism, then, can be blunted, it might be thought, by appeal (1) to the non-perversity of morality, and (2) to the extensional equivalence of the morally required and that which no morally conscientious person would knowingly fail to do.
5. Actualism’s Difficulties with Beneficial Sacrifice
The Actualist may indeed have a passable reply to the charge that Actualism implausibly allows people to get out of having certain present moral obligations simply in virtue of their potential future moral failings. Possibilism is most certainly not the dominant view. In fact, the Actualism/Possibilism debate has settled into a seemingly intractable stalemate. I hope to break through this stalemate. I argue that Actualism has difficulties with a particular moral phenomenon, one I call “permissible beneficial sacrifice”, difficulties which cannot be circumvented by the kinds of replies Actualists offer in response to the standard challenge to Actualism. First, I distinguish between the two dominant versions of Actualism currently in play, and then I show how they each run into trouble in cases of permissible beneficial sacrifice.
5.1. Two Actualisms
There are two main versions of Actualism. Both have at their core the following thesis about moral permissibility:
Actualism Core ( AC )
It is permissible for S to φ from t 1 – t 2 just in case (1) S is able, at t 1 , to φ from t 1 – t 2 ,23 24 and (2) there is no possible world that would be actual were S to do, from t 1 – t 2 , anything other than φ that S is able, at t 1 , to do from t 1 – t 2 which has a higher deontic value for S at t 1 25 than does the possible world that would be actual were S to φ from t 1 – t 2 .
Where the two versions of Actualism diverge is in their respective accounts of the notion of ability appealed to in AC . According to the first, what an agent is able to do at t 1 are all the actions and series of actions that she would perform were she to do each of the things she is able, at t 1 , to do at t 1 . According to the second version of Actualism, what an agent is able to do at t 1 are all the actions and series of actions that she would perform were she to do each of the things she is able, at t 1 , to do at t 1 and also all of the actions and series of actions she would later be able to perform were she to do each of those things.26 More precisely, the account of ability appealed to in AC according to the first version of Actualism is:
| Able(1): | S is able, at t 1 , to φ from t 1 – t 2 iff there is some ψ such that | |
| (i) S is able, at t 1 , to ψ at t 1 , and | ||
| (ii) were S to ψ at t 1 , it would be the case that S φs at t 2 . | ||
And the account of ability appealed to in AC according to the second version of Actualism is:
| Able(2): | S is able, at t 1 , to φ from t 1 – t 2 iff there is some ψ such that | |
| (i) S is able, at t 1 , to ψ at t 1 , and | ||
| (ii) were S to ψ at t 1 , it would be the case that S is able, at t 2 , to φ at t 2 .27 | ||
Henceforth, let ‘able1’ refer to the notion given by Able(1) and let ‘able2’ refer to that given by Able(2). The first version of Actualism, Actualism 1 , is that which is gotten by replacing all occurrences of ‘able’ in AC with ‘able1’, and the second version of Actualism, Actualism 2 , is that which is gotten be replacing all such occurrences with ‘able2’.28 29
Consider Headache once again. Here are the relevant ability data in that case:
-
On Monday, Doctor is both able1 and able2 either to give d or to ~give d on Monday .
-
On Monday, Doctor is not able1, but is able2, to [give d & give e] from Monday to Tuesday .
Given these data, Actualism 1 and Actualism 2 agree about what it is permissible for Doctor to do on Monday—namely, ~give d, and only ~give d. But they disagree about what it is permissible for Doctor to do from Monday to Tuesday. Actualism 1 dictates that it is permissible only for Doctor to [~give d & ~give e] from Monday to Tuesday; and this is because Doctor is able1 only to [~give d & ~give e] or to [give d & ~give e] from Monday to Tuesday, and of those two options [~give d & ~give e] is clearly best. Actualism 2 dictates that it is permissible only for Doctor to [give d & give e] from Monday to Tuesday; and this is because Doctor is able2 to [give d & give e] from Monday to Tuesday and that is clearly her best option from Monday to Tuesday.
Because Actualism 2 has it that Doctor both is not morally obliged to give d on Monday and also is morally obliged to [give d & give e] from Monday to Tuesday it is thus committed to rejecting a popular deontic principle:
‘And’-Elimination for Moral Obligation (AEMO): Necessarily, if S is morally obliged to (φ & ψ) from t φ - t ψ (t φ ≤ t ψ ), and S can, at t φ , φ at t φ then S is morally obliged to φ at t φ .30
To many, this is a huge cost; it seems counterintuitive that one could be obliged to carry out some course of action but not also be obliged to do that which is the necessary first step of carrying it out. Many find the cost of giving up AEMO so great as to render Actualism 2 a non-starter. (I side with the too-great-a-cost camp.) However, one benefit of Actualism 2 , it might be suggested, is that it allows us to say that though it would be morally impermissible for Doctor to give d on Monday, nevertheless, if Doctor does ~give d on Monday, that will entail her failing to [give d & give e] from Monday to Tuesday, thereby falling short of fulfilling all of her moral obligations.31 And there does seem to be something morally deficient in Doctor when she fails to give Patient drug D on Monday in Headache. She’s knowingly refraining from performing the course of action which would completely cure Patient.
This purported benefit of Actualism 2 over Actualism 1 is a mirage. Any intuitive thought of Doctor’s exhibiting a moral failing in Headache can be easily accounted for by the proponent of Actualism 1 by noting that Doctor, in virtue of her current disposition to fail to act on Tuesday, is seriously morally deficient. (This is important because, for there to be a sense of moral failing on the part of Doctor on Monday, the proponent of Actualism 1 can plausibly maintain, her potential future failure to act must reflect something about Doctor’s character on Monday. Were she not disposed on Monday to not administer E on Tuesday, then there wouldn’t be any intuitive sense of moral failing on the part of Doctor on Monday in need of being given an account of.32) Being disposed not to do that which one could easily do to save someone else’s life is a serious moral character flaw. What’s more, this explanation may well be better than that offered by the proponent of Actualism 2 . For in versions of Headache in which Doctor’s potential future failure to administer E would not be caused by a moral character flaw, Actualism 1 needn’t provide grounds for any intuitive thought that there is any such moral failing.
Headache (Switched Labels) : Everything is as it is in Headache except that were Doctor to administer D on Monday, she would fail to administer E on Tuesday, not because of her own potential future laziness, but because she knows, on Monday, that on Tuesday the labels on the drugs, unbeknownst to her then, will have been switched around (the result of the dastardly behavior of some evil third party). Doctor does not administer D on Monday.
Here, it may seem, there isn’t any moral failing on Doctor’s part that needs accounting for at all, and yet the proponent of Actualism 2 is committed to the claim that in doing ~give d on Monday, Doctor fails to carry out one of her obligations, namely, to [give d & give e] from Monday to Tuesday. And she claims that that kind of failure grounds an intuitive thought that there is some moral failing exemplified in Doctor’s doing ~give d on Monday in Headache. But, and here’s the problem, there is no such intuitive thought regarding Headache (Switched Labels).33
None of this is to say, of course, that Actualism 2 doesn’t have advantages over Actualism 1 . It’s only to say that its being able to accommodate a certain sense that there is some moral failing associated with Doctor’s doing ~give d on Monday in Headache isn’t one of them. Though I do not think Actualism 2 has any advantages over Actualism 1 , and Actualism 1 has the merit of not entailing the falsity of AEMO, both versions of Actualism have serious problems with merely permissible beneficial sacrifice.
5.2 Actualism and Beneficial Sacrifice
Having thus distinguished the two main versions of Actualism, I now show how they each have trouble with merely permissible beneficial sacrifice.
First, a couple words about beneficial sacrifice. By ‘beneficial sacrifice’ I mean any action whereby the agent imposes a loss upon, or allows a loss to befall, someone—either the agent herself (an instance of beneficial self-sacrifice) or someone else (an instance of beneficial other-sacrifice)—in order to prevent yet others from suffering a loss. Some beneficial sacrifice is morally required. If I can save an innocent person from drowning but only by doing something that will cause me, or even you, a two-hour headache, then I’m morally required to save the person and cause the headache. Some beneficial sacrifice is morally impermissible, of course—as when I kill you in order to prevent someone else from suffering a two-hour headache. And, finally, some beneficial sacrifice is neither impermissible nor morally required, it is merely morally permissible. If I can save a person from drowning, but only at the cost of my own life, then I’m morally permitted, though not required, to sacrifice my life to save the person. And if I can save a person from drowning, but only by doing something that will cause you the loss of two fingers, then I’m morally permitted, though not required, to save the person at the cost to you of your two fingers.34
Merely permissible beneficial sacrifice makes trouble for both versions of Actualism. I begin by considering their troubles with merely permissible beneficial self-sacrifice.35 Not only do both versions of Actualism allow people to get out of having certain present moral obligations in virtue of their potential future moral failings, they also, much more troublingly, allow people to get out of having certain present moral obligations in virtue of their potential future morally supererogatory behavior.
Take Actualism 1 first and consider the following case involving diachronic supererogation:
Three Drowning : Chang stands at the edge of a pond in which three people are drowning. At t 1 , Chang’s options are:
dive: dive in and save two of the three~dive: not dive in and allow all three to drown
If Chang were to dive at t 1 , then her options at t 2 would be:
stay: stay in the pond and save the remaining person at the cost of her leg (a crocodile would bite off Chang’s leg if she stays and saves the one)exit: exit the pond, thereby allowing the remaining person to drown and not losing a leg
Chang knows at t 1 that if she were to dive at t 1 , then, though she would be able to exit at t 2 , because she would then be so committed to saving as many people as she can, she would freely (non-pathologically, non-compulsively, etc.) and fully-informedly choose to stay, thereby saving the remaining person and losing her leg. (Chang also knows at t 1 that if she were to ~dive at t 1 , then at no time after t 1 would she be able to save any of the three, and so all three would drown.)
In Three Drowning the relevant ability data are:
-
At t 1 , Chang is both able1 and able2 either to dive or to ~dive at t 1 .
-
At t 1 , Chang is not able1, but is able2, to [dive & exit] from t 1 to t 2 .
Because at t 1 Chang is only able1 to dive or ~dive at t 1 , Actualism 1 deems it morally permissible for Chang to ~dive at t 1 in Three Drowning. And that’s because were Chang to dive at t 1 she would stay at t 2 , thereby losing a leg and saving three people, and no one is morally required to lose a leg in order to save one or even three people from drowning.36 But it is not morally permissible for Chang to ~dive at t 1 in Three Drowning. Chang is not permitted, for any ordinary and intuitive sense of ‘permitted’, to allow the two people she could easily save, at no cost to herself, from drowning at t 1 to drown.
It is important to keep in mind when considering Three Drowning that Chang’s potential future rescuing of the third person at the cost of her leg at t 2 were she to dive at t 1 is something that she would freely, uncoercedly, and uncompelledly do at t 2 . Chang’s knowledge that she would do this were she to dive at t 1 is just the everyday, ordinary kind of knowledge that a procrastinator has of her own future voluntary procrastination; the procrastinator knows ahead of time that she will freely, uncoercedly, and uncompelledly procrastinate in the future.37 We mustn’t think that because Chang knows that she will come to want, and act so as, to save as many people as she can at t 2 were she to dive at t 1 that this means that there is any sense in which at t 1 she is already committed to or in the grip of some general deep standing commitment to save as many people as she can. Again, she simply knows, perhaps from past experience—just as the procrastinator can know of her potential future procrastination from past experience—that she will end up going above and beyond the call of duty and saving the third person at the cost of her leg at t 2 were she to dive at t 1 . All the same, the fact that she knows this in no way makes it okay now for her to fail to rescue the two she could easily save at t 1 .
In response to all of this the proponent of Actualism 1 might suggest that its being permissible for Chang to ~dive at t 1 isn’t actually counterintuitive at all. After all, she might say, there is nothing Chang is able at t1 to do such that were she to do it she would rescue the two without losing her leg, and no one is morally obliged to lose a leg in order to save two, or even three, people from drowning. Here the proponent of Actualism 1 is leaning heavily on the thought that, given the set-up of the case, Chang simply can’t, she’s simply unable to, rescue the two without losing her leg. But that’s just to assume that Able(1), instead of Able(2), is the morally relevant notion of cross-temporal ability, and that’s precisely what’s at issue between Actualism 1 and Possibilism.38 If Able(2) is the morally relevant notion of cross-temporal ability, then Chang is able at t 1 , in the morally relevant sense of ‘able’, to rescue the two without losing her leg. To adjudicate between Possibilism and Actualism 1 , then, we need to approach Three Drowning without assuming either Able(1) or Able(2) is the morally relevant notion of cross-temporal ability. Once we do this, however, the proponent of Actualism 1 ’s case for the permissibility of Chang’s not diving in and saving the two at t 1 falls apart.
Consider the following conversation between Chang and Torres, an on-looker, after Chang refrains from diving in and saving the two at t 1 :
| Torres: | Why didn’t you dive in and save those two people?! | |
| Chang: | It was okay for me not to. | |
| Torres: | What?! Of course it wasn’t okay. You could have dived in and then you could have easily lifted them out. | |
| Chang: | No. You see there’s a crocodile in there and if I saved them it would’ve bitten my leg off. | |
| Torres: | I see. If you jumped in and saved them you’d not have been able to avoid losing your leg. | |
| Chang: | Oh no. I didn’t say that. If I had dived in and saved them I would have quite easily been able to get out without losing my leg. It’s just that I know myself: if I had dived in, I would have chosen to stay in and because of that choice I would make, the crocodile would bite my leg off. | |
Chang’s explanation of why she’s permitted to let the two people drown falls flat. Her defense would be a lot stronger if she could somehow sneak in that she can’t save the two without losing her leg. But, again, that would simply be assuming that Able(1), and not Able(2), is the morally relevant notion of cross-temporal ability, something to which, in this dialectical context, she is not entitled.39 Shorn of an appeal to inability, Chang’s defense of her not saving the two crumbles. And that it crumbles puts the lie to the proponent of Actualism 1 ’s claim that its being permissible for Chang to ~dive at t 1 isn’t counterintuitive.
Three Drowning poses a more severe problem for Actualism 1 than that posed by Headache. In Headache, the proponent of Actualism 1 can support her claim that Doctor is not morally required to give d on Monday in Headache by pointing to the fact that were she to give d on Monday, that would lead to her acting more morally wrongly overall than she would were she to ~give d on the supposition that give d is required. Nothing like this is available to the proponent of Actualism 1 to explain how Chang could be permitted not to save the two people she could easily save at t 1 in Three Drowning. If she can easily save the two at t 1 and afterwards easily avoid any injury to herself by getting out of the pond, which it would then be perfectly permissible for her to do, how could she possibly be permitted to not save the two at t 1 ? Likewise, in Headache, the fact that no person who is being morally conscientious on Monday in Headache would give d on Monday might support the Actualist’s claim that in Headache it is morally impermissible for Doctor to give d on Monday. But if that’s so, then the fact that no morally conscientious person would fail to dive in and save the two at t 1 in Three Drowning—as no morally conscientious person would fail to do—should also establish that it is morally impermissible, and not permissible, as Actualism 1 dictates, for Chang to ~dive at t 1 in Three Drowning. None of the defenses proponents of Actualism offer up in response to the kinds of Headache-based criticisms leveled against it are available in response to the Three Drowning-based objection to Actualism 1 .
Actualism 2 might seem to be able to avoid the problem posed by Three Drowning. Like Actualism 1 , and for the very same reason, Actualism 2 has the consequence that it is morally permissible for Chang to ~dive at t 1 . However, unlike Actualism 1 , Actualism 2 also has the consequence that it is not morally permissible for Chang to ~dive from t1 to t2.40 That’s because Chang is able2, at t 1 , to [dive & exit] from t 1 to t 2 and if one can [dive & exit], instead of ~dive, from t 1 to t 2 , one morally ought not to ~dive from t 1 to t 2 . ( Actualism 1 does not have this consequence because Chang is not able1, at t 1 , to [dive & exit] from t 1 to t 2 in Three Drowning.) Actualism 2 , then, may be thought to have resources Actualism 1 lacks in responding to the Three Drowning-based objection. Though Actualism 2 does yield that it is permissible for Chang to ~dive at t 1 , it can nonetheless offer a relatively straightforward error theory for the thought that it is impermissible for Chang to ~dive at t 1 : it seems impermissible for Chang to ~dive at t 1 , the proponent of Actualism 2 might contend, because there is something close to it, namely, Chang’s doing ~dive from t1 to t2, which is impermissible for Chang. Mightn’t this allow Actualism 2 to avoid the objection from supererogation?
No. What is objectionable about Actualism 1 is not merely that it can’t somehow account for the thought that it is morally impermissible for Chang to allow the two people she could save to drown, it’s that it allows that there is a sense in which it is permissible for Chang to let them drown. Intuitively, however, there is no sense in which it is permissible—morally ok, morally allowed, morally fine, etc.—for Chang to let the two she is able (both able1 and able2) to easily save drown. Since Actualism 2 has this consequence, the Three Drowning-based objection tells against Actualism 2 just as much as it does against Actualism 1 .41
Actualism has trouble not only with merely permissible beneficial self-sacrifice, but with merely permissible beneficial other-sacrifice as well. Consider the following case:
Trolley : Six people are trapped on a track along which an out-of-control trolley is hurtling. At t 1 , a bystander, Khan, has two options:
pull: pull four of the six off the track, thereby saving them~pull: not pull the four off the track, thereby allowing all six to be run over and killed
If Khan were to pull at t 1 , her doing so would enable her at t 2 to flip a switch which would turn the trolley away from the remaining two and onto a side-spur of track on which one other innocent person is trapped. (If she doesn’t pull at t 1 she won’t be able to flip the switch at t 2 because the four would be standing in the way of her reaching the switch.) And so, if Khan were to pull at t 1 , her options at t 2 would be:
turn: turn the trolley away from the remaining two, thereby saving them, and onto the one, thereby killing him~turn: not turn the trolley, thereby allowing it to run over the remaining two and not killing the one
Khan knows at t 1 that if she were to pull at t 1 , then, though she would be able to ~turn at t 2 , because she would then be so committed to saving as many people as she can, she would freely (non-pathologically, non-compulsively, etc.) and fully-informedly choose to turn, thereby saving the remaining two and killing the one.
In Trolley the relevant ability data are:
-
At t 1 , Khan is both able1 and able2 either to pull or to ~pull at t 1 .
-
At t 1 , Khan is not able1, but is able2, to [pull & ~turn] from t 1 to t 2 .
Because Khan is only able (both able1 and able2), at t 1 , to pull or ~pull at t 1 , both Actualism 1 and Actualism 2 deem it morally permissible for Khan to ~pull at t 1 in Trolley. And that’s because were Khan to pull at t 1 she would turn at t 2 , thereby killing an innocent person and saving six people. And though it is permissible to kill one in redirecting a threat away from two, it is not morally required to kill an innocent person, whether by redirection or not, in order to save two, or even six, people from being killed.42 But it is not morally permissible for Khan to ~pull at t 1 in Trolley. It is very seriously morally wrong for her to allow the four people she could easily pull off the track at t 1 to be run over and killed by the trolley. ( Actualism 2 also has the consequence that it is not morally permissible for Khan to ~pull from t 1 to t 2 in Trolley.43 But, as before, this is no saving grace, for allowing that there is some intuitive sense in which it is permissible for Khan to not pull the four off the track at t 1 is what’s objectionable.)
A proponent of either version of Actualism might try to respond to these objections by noting that in these cases were we to view the agent’s future self as morally equivalent to some third party, then she might well be able to deliver the intuitively correct verdicts in these cases. Certainly, if it were the case that Khan’s doing pull would have the result that some third party would turn the trolley away from the two and onto the one, we wouldn’t think that that fact would get Khan off the moral hook of having to save the four she can easily save. That’s as may be. But pointing to the intuitive verdicts in some other cases is not an adequate response to the Three Drowning- and Trolley-based objections to Actualism. The problem is Actualism’s implications in these cases; that it may get the right results in yet other cases doesn’t solve that problem.
The Actualist may of course tell a self-consistent story according to which she treats Three Drowning and Trolley as of a piece with cases in which it is a third party whose future action is relevant to the agent’s present moral obligations. But whether one’s own potential future voluntary actions should be treated as morally on a par with those of some other distinct person is precisely the question at issue between the Actualist and the Possibilist. What’s more, treating an agent’s future self as morally on a par with some other distinct person is out of step with much of the rest of our moral practice. For one example: if I know now, this morning, that later this afternoon I’m going to promise to bring cookies to the party tonight, and starting baking the cookies now, in the morning, is the only way I’ll be able to bring cookies to the party tonight, then I ought to, and insofar as I’m morally conscientious I will, start baking the cookies now; but were it some other person who was going to promise in the afternoon that she would bring cookies to the party tonight, it wouldn’t follow that I ought to start baking now, nor would it follow that moral conscientiousness entails my starting baking now (and this is true even if my starting baking now were the only way that other person would be able to bring the cookies tonight—we’re under no significant standing obligation, nor does moral conscientiousness require us, to ensure that others keep their promises). For another example: whereas I can now consent to your later punching me and thereby make it permissible for you to punch me later (this is, after all, what makes boxing a morally permissible activity), I cannot now consent to your later punching some third party and thereby make your later punching that third party morally permissible.44 These are just some of the many ways in which one’s future self is morally disanalogous to that of a distinct individual. One cannot simply appeal to the moral parity of one’s future self and some future distinct individual in replying to the Three Drowning- and Trolley-based objections to Actualism.
6. Conclusion
I’ve argued that Actualism and Subjective Possibilism are both untenable. It follows from this that whatever the correct moral theory is it must be an Objective Possibilist one.
Acknowledgments
I’d like to thank audiences at the University of Massachusetts Amherst and the Australian National University for helpful feedback on predecessors of this paper. I’d also like to thank Liz Harman, Alejandro Pérez-Carballo, and especially Chris Meacham, as well as two anonymous referees for Ergo and other anonymous referees at other journals, for feedback on and discussion of drafts of this paper.
References
Bykvist, Krister (2009). Objective versus Subjective Moral Oughts. Uppsala Philosophical Studies, 57, 39–65.
Cohen, Yishai and Travis Timmerman (2016). Moral Obligations: Actualist, Possibilist, or Hybridist? Australasian Journal of Philosophy, 94(4), 672–686.
Feldman, Fred (1986). Doing the Best We Can. D. Reidel.
Goble, Lou (1993). The Logic of Obligation, ‘Better’ and ‘Worse’. Philosophical Studies, 70(2), 133–163.
Goldman, Holly S. (1976). Dated Rightness and Moral Imperfection. Philosophical Review, 85(4), 449–487.
Goldman, Holly S. (1978). Doing the Best One Can. In Alvin I. Goldman and Jaegwon Kim (Eds.), Values and Morals (185–214). D. Reidel.
Graham, Peter A. (2010). In Defense of Objectivism about Moral Obligation. Ethics, 121(1), 88–115.
Greenspan, Patricia S. (1978). Oughts and Determinism: A Response to Goldman. Philosophical Review, 87(1), 77–83.
Humberstone, I. L. (1983). The Background of Circumstances. Pacific Philosophical Quarterly, 64, 19–34.
Jackson, Frank (1991). Decision-Theoretic Consequentialism and the Nearest and Dearest Objection. Ethics, 101(3), 461–482.
Jackson, Frank (2014). Procrastinate Revisited. Pacific Philosophical Quarterly, 95(4), 634–647.
Jackson, Frank and Robert Pargetter (1986). Oughts, Options, and Actualism. Philosophical Review, 95(2), 233–55.
Kiesewetter, Benjamin (2011). ‘Ought’ and the Perspective of the Agent. Journal of Ethics and Social Philosophy, 5(3), 1–24.
Moore, George E. (1903). Principia Ethica. Cambridge University Press.
Portmore, Douglas W. (2011). Commonsense Consequentialism: Wherein Morality Meets Rationality. Oxford University Press.
Prichard, Harold A. (1949). Duty and Ignorance of Fact. In H. A. Prichard Moral Obligation (18–39). Clarendon Press.
Ross, William D. (1939). Foundations of Ethics. Oxford University Press.
Ross, William D. (2002). The Right and the Good. Oxford University Press.
Ross, Jacob (2012). Actualism, Possibilism, and Beyond. In Mark Timmons (Ed.), Oxford Studies in Normative Ethics (Vol. 2, 74–96). Oxford University Press.
Sidgwick, Henry (1907). The Methods of Ethics. Macmillan.
Smith, Holly M. (2014). The Subjective Moral Duty to Inform Oneself before Acting. Ethics, 125(1), 1–28.
Sobel, Jordan Howard (1976). Utilitarianism and Past and Future Mistakes. Noûs, 10(2), 195–219.
Thomason, Richmond H. (1981). Deontic Logic and the Role of Freedom in Moral Deliberation. In Risto Hilpinen (Ed.) New Studies in Deontic Logic (177–186). D. Reidel.
Thomson, Judith J. (1990). The Realm of Rights. Harvard University Press.
Vessel, Jean-Paul (2009). Defending a Possibilist Insight in Consequentialist Thought. Philosophical Studies, 142(2), 183–195.
Vessel, Jean-Paul (2015). Against Securitism, the New Breed of Actualism in Consequentialist Thought. Utilitas, 28(2), 164–178.
Vorobej, Mark (2000). Prosaic Possibilism. Philosophical Studies, 97(2), 131–136.
Woodard, Christopher (2009). What’s Wrong With Possibilism. Analysis, 69(2), 219–226.
Zimmerman, Michael J. (1996). The Concept of Moral Obligation. Cambridge University Press.
Zimmerman, Michael J. (2008). Living with Uncertainty. Cambridge University Press.
Zimmerman, Michael J. (2014). Ignorance and Moral Obligation. Oxford University Press.
| Able(2)′: | S is able, at t 1 , to φ from t 1 – t 2 iff there is some ordered series of compossible act tokens <ψ 1 ,…,ψ n > such that |
| (i) S is able, at t 1 , to ψ 1 at t 1 , | |
| (ii) were S to perform <ψ 1 ,…,ψ n >, it would be the case that S is able, at t 2 , to φ at t 2 . | |
| (iii) for each nesting subseries, <ψ 1 ,…,ψ i+1 >, of <ψ 1 ,…,ψ n >, were S to perform ψ 1 & … & ψ i , it would be the case that S is able, at t i+1 , to ψ i+1 at t i+1 . |
The problem for Possibilism is that, plausibly, Tran is not morally obligated to push the button in Button, but it can seem hard to see how Possibilism can deliver this verdict. Tran can, after all, prevent Ito’s breaking his leg while at the same time not being tortured for 1000 years, and so, from the Possibilist point of view, if she can do those things, then presumably she is morally obliged to push the button. Appearances notwithstanding, Button does not pose a problem for Possibilism, and that’s because it can in fact deliver the verdict that Tran is not morally obligated to push the button. What Button does show, I believe, is only that the moral permissibility of an agent’s action may indeed partly supervene on her subjective state. (Note that this is consistent with my definition of Objectivism; see Footnote 4.) What makes it the case that Tran is not morally obligated to push the button at t 1 in Button is that the harm of suffering 1000 years of torture is not morally relevantly avoidable for her at t 1 . Morally relevantly avoidable harm clearly has a subjective component: if you throw a baseball at the back of my head, though I could have avoided it by stepping aside at the moment it was about to hit me, because I had no idea that you were throwing it at me, when it hit me in the back of the head, that harm was not a morally relevantly avoidable one for me. So too, though there is some precise sequence of actions Tran could take after pressing the button such that she would be able to avoid being tortured for 1000 years were she to push the button and then perform that sequence of actions, because she wouldn’t know at every moment throughout her 80-year life how she’d have to behave to avoid being tortured for 1000 years, the torture were she to suffer it would not be morally relevantly avoidable for her when she suffers it. (Were it stipulated, on the other hand, that she knew at t 1 that at every moment in her life she would be aware of which choice would lead to her being tortured for 1000 years, then though the Possibilist would then have to say that she is morally obliged to push the button at t 1 , that would be the intuitive verdict in such a case.) And as she is morally obligated to save Ito from suffering a broken leg at t 1 only if any harm greater than that which would be supererogatory for preventing a broken leg she might suffer in preventing it is such that were she to prevent Ito’s breaking his leg either she wouldn’t suffer it or her suffering it would have been morally relevantly avoidable for her when she suffered it, it is not the case, even for the Possibilist, that Tran is morally obligated to push the button at t 1 in Button. That it is the subjective nature of the avoidability of the harm in question which is at the root of Tran’s not being morally obligated to push the button can be seen by noticing that if in Button it were in addition stipulated that Tran knows at t 1 that were she to press the button she in fact wouldn’t perform a morally wrong action throughout the rest of her 80-year life, then, intuitively, it would be the case that she is morally obligated to push the button at t 1 . ⮭Button : Ito is about to fall and break his leg. The Devil presents Tran with a button and says, “If you push this button at t 1 , I will prevent Ito from breaking his leg; but if you push the button at t 1 and then freely perform one wrong action (no matter how trivial) for the rest of your (80-year) life, you will be tortured for 1000 years.”