Instrumental Reasons Chapter 2

Chapter 2 But Super-Strong Availability is still more restrictive. It writes off further cases of instrumental transmission: namely, all those in which the reason to take the means is to help to put oneself in a position, at some later time, from which there is something that one can (but might not) do to guarantee the end. (Indeed, for Raz 2005a, this is by definition what “facilitating plans” do.) Moreover, there are counterexamples that conform even to Super-Strong Availability. In order to save his arm, Mountaineer must now cross over a bridge to the other side of a chasm. But if he steps now, he is almost certain to look down while doing so. And if he looks down, he will almost certainly not successfully cross. Indeed, he will almost certainly lose his balance and plummet to his death. All the same, it is entirely within his control to step without looking down. He doesn’t know, although we know, that stepping while looking down will upset his balance. Stepping without looking down now is super-strongly available, and he ought to step without looking down now. A necessary means to stepping without looking down now is stepping now, which is also super-strongly available. But it is not the case that he ought to step. Indeed, he ought not to step, since it will almost certainly be his death. Reply 3: A final reply to these alleged counterexamples to Ought Necessity is to grant that Unlucky, Procrastinate, and Mountaineer ought to achieve the end, but to argue that they ought also to take the necessary means. As for Cost-Unlucky, the reply would grant that he ought not take the necessary means, but would argue that it is also not the case that he ought to achieve the end. Since Refrain Reasons-Ought and Relevant Reasons Ought, as we have seen, support the opposite conclusions, this approach needs some other account of the relation between reasons and oughts. The likely candidate would be: Necessary Reasons-Ought: One ought to X iff X-ing is, or is a necessary means to, what, from among all the available options one has most reason to do. Suppose that the option for which Unlucky has most reason includes saving as a necessary component. This will be some “exhaustive” option, which determines, for every available action that might relevantly affect the outcome, whether or not one performs that action, such as: saving the antique and donating it to Oxfam and working tirelessly henceforth to end world hunger and so on. But then it will also be true, on this account, that Unlucky ought to taxi as well. Similarly for Procrastinate and Mountaineer. For his part, Cost-Unlucky can’t have most reason for an option that includes his saving the antique, since that will always be a net loss. So it follows that it is not the case that he ought to save it. At this point, intuitions may simply diverge. But mine incline me to affirm both that it is not the case that Unlucky/ Procrastinate/ Mountaineer ought to taxi/accept/step and, moreover, that Unlucky/ Procrastinate/ Mountaineer ought not to taxi/ accept/ step. (I set aside CostUnlucky for simplicity.) Among other reasons to affirm these claims are these four: (A) It seems to me counterintuitive that one ought to do something that, if one does it, will be a disaster, as compared with other options available to one. (B) It seems to me counterintuitive that one ought to do something that one has more reason to refrain from doing. Most importantly, these claims seem to me to track (C) good and bad advice, and (D) good or bad deliberative conclusions, at least in the cases of Unlucky and Procrastinate. Suppose all the facts are laid out for Unlucky or Procrastinate, and they ask you, or ask themselves, “So, knowing what we know, what ought I to do now?” It seems bad advice to say, “You ought to taxi/accept,” and good advice to say, “You ought not taxi/accept. Given that you are so unlikely to save/write, it will be a disaster if you do. You have more reason to refrain.” For the same reasons, it seems a bad deliberative conclusion for them to reach themselves, “So I ought to taxi/accept.” The correct conclusion, it seems, is to say, “But I ought not taxi/accept. Given that I am so unlikely to save/write, it will be a disaster if I do. I have more reason to refrain.”27 One might propose theories of error for these intuitions. For instance, one might stress the distinction between something that one ought to do and all that one ought to do. To say, “He ought to taxi/accept/step,” while strictly true, is misleading, since it suggests that this is all that he ought to do. A fuller answer would be: “He ought to save/accept and write/step without looking down, and that requires his taxi-ing/accepting/stepping. That is why he ought to taxi/accept/step.” Similarly, it sounds odd to say that one ought to do what will be a disaster if one does it, or that one ought to do something that one has more reason to refrain from doing, only because “ought” is here heard as “all that one ought,” and, so heard, the claims are actually false. It won’t be a disaster if one does all that one ought, and one does not have more reason to refrain from doing all that one ought. Alternatively, with Kiesewetter ms., one might suggest that the intuitions confuse a “narrow-scope” with a “wide-scope” ‘ought.’ It is false that if he is unlikely to write/avoid looking down, then he ought accept/step. What is true is that he ought to see to it that (if he is unlikely to write/avoid looking down, then he does not accept/step). These theories of error might account for (A) and (B). But they don’t account for (C) and (D). In the episodes of advice and deliberation that we described, all the facts are made plain to all parties. So there can be no lingering confusion between “some of what” and “all of what” one ought to do. And on the wide-scope view, we can never get to the desired answer to the question, “So, what ought I to do now?” The closest we can come is the unhelpful answer, “You ought either to be likely to write in the future, or not to accept.” One way of dealing with (C) and (D) is to revert to Super-Strong Availability, so that Mountaineer is left as the only relevant case. The advice, “You ought to step,” and the conclusion, “I ought to step,” seem fine in the Mountaineer’s case. For once we give the Mountaineer the full story, or he gives it to himself, he comes to know that he ought to step without looking down if he steps. And so it becomes very likely that if he steps, he will step without looking down. And so “You ought to step” becomes perfectly good advice, and “I ought to step” becomes a perfectly good deliberative conclusion. Apart from the worry that SuperStrong Availability is too restrictive, therefore, I have no further argument against this reply. Another way of dealing with (C) and (D) is to deny, on the one hand, that the badness of advice is persuasive evidence of its falsity and, on the other, that the goodness of advice is persuasive evidence of its truth. What makes advice good or bad is, instead, whether one ought to or ought not to give it. Suppose that if X is now under the agent’s control, and if we say to the agent, “You ought to X,” then the agent will X. However, if X is not now under the agent’s control, then saying to the agent, “You ought to X,” will have no effect. What, then, ought we to say to, say, Procrastinate? Clearly, we ought not to say, “You ought to accept,” because this will lead Procrastinate to accept, but make him no more likely to write, if he accepts. Indeed, we ought to say, “It is not the case that you ought to accept.” This is the sense in which “You ought accept” is bad advice, but “It is not the case that you ought accept” is good advice. The adviser ought not say the former and ought to say the latter. All the same, this bad advice is true, and this good advice is false. The first problem with this proposed explanation is that it lets truth and falsity diverge too far from the goodness or badness of the corresponding advice. To be fair, even an opponent of Ought Necessity must accept some divergence. In particular, he must agree with the proponent of Ought Necessity that some true ‘ought’ claims will be bad advice. Typically, in advising someone to X, we are aiming to lead that person to X by way of leading him to decide to X. But presumably we ought not lead Procrastinate to decide to accept and write, since that is apt to lead Procrastinate to accept. For that reason, telling Procrastinate, “You ought to accept and write,” seems like bad advice, even though the opponent of Ought Necessity, no less than the present proponent, agrees that this claim is true.28 However, the defender of Ought Necessity is arguing not only that some true “ought” claims are bad advice, but moreover that some false “ought” claims, such as “You ought not accept,” are good advice. It’s more surprising that a good adviser sometimes ought to keep his counsel about what the acknowledged facts imply than that he ought to lie to his advisees about it. The second problem is that this explanation doesn’t work in the case of first-person deliberation. Applied to that case, the idea would be to explain why “I ought to accept” is a bad deliberative conclusion by appeal to the thesis that a deliberative conclusion is bad if one ought not to judge it, but good if one ought to judge it. But suppose that Procrastinate will accept only if he concludes, “I ought to accept”—not an unlikely situation. Then, according to Necessary Reasons-Ought, he ought to conclude, “I ought to accept.” But then “I ought to accept” should be a good deliberative conclusion, which is the opposite of what was to be explained. 4. Second desideratum: means must not be superfluous To recap: We have identified one desideratum for a principle of instrumental transmission: namely, Means Probabilize. Then we observed that it tells against Strong Necessity and, arguably, also against Ought Necessity. Now we turn to the second desideratum, which is revealed by a defect in Means Probabilize, although a defect that does not affect the arguments given so far. Suppose Dr. Twoways has the end of relieving the patient’s pain. Drug 1 alone will do this for sure. If he also gives Drug 2, it will at first neutralize Drug 1, but then combine with Drug 1 in the patient’s bloodstream to become Drug 3, which will also relieve the patient’s pain for sure. Suppose that at every possible outcome Twoways gives the patient Drug 1. (Either it is known, or fated, that he will give Drug 1.) In this case, it seems that no reason transmits to Twoways’s giving Drug 2. However, Means Probabilize implies that reason does transmit. The probability, conditional on giving Drug 2, that Twoways relieves the patient’s pain and that giving Drug 2 helps to bring this about, is one. In every possible outcome in which he gives Drug 2, his doing so helps to cause the relief of pain. In every such outcome, Drug 3 is the proximate cause of the relief of pain, and the presence of Drug 3 is a causal consequence of having administered Drug 2. The reason why no reason transmits, one wants to say, is that in every possible outcome in which Drug 2 helps to bring it about that the pain is relieved, Drug 2 brings this about superfluously. In other words, we should replace Means Probabilize with: Means Probabilize Nonsuperfluously: If there is reason for one to E, and there is positive probability, conditional on one’s M-ing, that one’s M-ing, or some part of one’s M-ing, helps to bring it about that one E’s in a nonsuperfluous way, then there is reason for one to M, whose strength depends on the reason for one to E and on the probability. The intuitive idea is that one’s M-ing is superfluous with respect to one’s E-ing in outcome W just when, if one’s M-ing were “removed” from W, one would still E. Consider any outcome in which Twoways gives Drug 2. If we “took away” Twoways’s giving Drug 2 from that outcome, Twoways’s giving Drug 1 would “still be there,” and so Twoways would still relieve the patient’s pain. To say “if one’s M-ing were ‘removed’ from W, one would still E” is not to say “if one did not M, one would still E.” One difference is that the counterfactually “closest” world, as it were, in which one does not M may be one in which not only is one’s M-ing “taken away,” but also something new is “added in.” For example, suppose the end is entering the house. In every possible outcome in which I enter by the front door, it is true that if I had not entered by the front door, I would have entered by the back door. In other words, at the “closest” world at which I do not enter by the front, not only is my entering by the front “taken away,” but also something new, my entering by the back, is “added in.” Intuitively, we don’t want to count entering by the front door as superfluous. If we did, then no reason would transmit to entering by the front door, and (if the situation is symmetrical) no reason would transmit to entering by the back door, even though (entering either by the front or by the back) would be necessary and sufficient for the end. But it would surely be wrong to say of me, after I enter by the front, not only that there was no more reason for me to do that than to enter by the back (which is true enough), but moreover that, as far as entering the house was concerned, there was no reason for me at all to do it—as one might properly say that I had no reason to tie myself to a stake in the lawn, preventing myself from entering the house. Another reason why “if one’s M-ing were ‘removed’ from W, one would still E” is not equivalent to “if one did not M, one would still E” is that the “closest” world in which one does not M may be one in which only part of M-ing, not the whole of it, is “taken away.” For example, it might be the case at every outcome at which Twoways (gives Drug 1 and gives Drug 2) that, if Twoways did not give both, Twoways would still give Drug 1. Even though, given that Twoways gives Drug 1, giving Drug 2 is superfluous, we don’t want to say that given that Twoways gives Drug 1, giving both is superfluous. Suppose that giving both drugs (and only giving both) will improve the patient’s complexion, and that giving both is no more costly than giving Drug 1 alone, since Drug 2 is about to spoil anyway. Intuitively, Twoways has more reason to give both than to give only Drug 1: that will relieve the patient’s pain and, without any additional cost, improve his complexion. But if we count giving both as superfluous, then reason to relieve pain will transmit only to giving Drug 1 alone. Since reason to relieve the patient’s pain outweighs the reason to improve the patient’s complexion, Twoways will have more reason to give only Drug 1 than to give both. To capture the intuitive idea, therefore, we should say something like this: Definition of superfluous means: One’s M-ing is superfluous with respect to one’s E-ing in an outcome W iff there is some M* such that: (i) one M*’s in W (ii) one’s M*-ing in W is not a part of one’s M-ing at W (iii) at W it is true that, if one did not M, then one would still M* in the same way and one would E and one’s M*-ing would help to bring about one’s E-ing. Clause (i) saves entering by the front door from superfluity, because when one enters by the front door, one does not also enter by the back door.29 Clause (ii) saves giving both drugs from superfluity, on the assumption that giving Drug 1 is part of (giving Drug 1 and giving Drug 2). It is crucial to distinguish the notion of a superfluous means from the notion of an “excessively costly” means. An excessively costly means is a means that is more costly than an alternative means that is at least as good a means.30 Reason does transmit to excessively costly means. It is just that the reason transmitted is typically outweighed by the excessive cost. While giving both drugs is not superfluous, for instance, it may well be excessively costly. For example, it may use up more scarce medicine than giving Drug 1 alone, which is at least as good at relieving the patient’s pain. While there is as much reason for giving both as for giving Drug 1 alone—either option is equally good a means to relieving the patient’s pain—there is stronger reason to refrain from giving both—it uses up more medicine.31 5. Why the Sufficiency principles have trouble with the nonsuperfluity desideratum If Means Probabilize Nonsuperflously is correct, then it poses a problem for the Sufficiency principles. The Sufficiency principles may seem plausible, because if one takes some sufficient means, then one is sure to achieve the end. The problem is that compatibly with that truth, one may be sure to achieve the end even if one does not take that sufficient means. This can happen, in particular, because the sufficient means are superfluous. In such cases, at least if Means Probabilize Nonsuperfluously is on the right lines, no reason transmits to those sufficient means. To make the case more exactly, let’s use an analogue to our earlier definition of necessary means: Definition of Sufficient Means: One’s M-ing is a sufficient means to one’s E-ing iff (i) in every possible outcome in which one M’s, one’s M-ing helps to bring about one’s E-ing, and (ii) there is some possible outcome in which one M’s. So understood, a sufficient means can be superfluous in every possible outcome at which it is taken. Indeed, this is exactly what happens with Dr. Twoways. Drug 2 is a sufficient means to relieving the patient’s pain. So, according to Weak Sufficiency, Twoways has reason to give Drug 2. However, since Drug 2 is superfluous toward the end of relieving pain in every possible outcome, Twoways intuitively has no reason to give Drug 2. And if Weak Sufficiency fails,32 then so does Strong Sufficiency. If no reason transmits, then a fortiori reason at least as strong does not transmit. If we assume Refrain Reasons-Ought or Relevant Reasons-Ought, then Ought Sufficiency also fails. Suppose that there is some reason to refrain from giving Drug 2. Then, since there is no reason for giving Drug 2, and some reason to refrain from giving it, Refrain Reasons-Ought implies that it is not the case that Twoways ought to give Drug 2. However, the untenability of Ought Sufficiency is not very surprising. It is also falsified, in less exotic cases, by sufficient means that are excessively costly, even if not superfluous. Suppose we ought to visit my folks for Thanksgiving. Options A and B are sufficient means in the sense defined. But A involves a 36-hour, $5,000 flight with six layovers, where B involves only a six-hour, $150, direct flight. Presumably, it is not the case that we ought to take A. One might object to this argument against Weak Sufficiency, on the grounds of: Reasons Non-Necessity: One has reason to X only if it is possible that one does not X. Since is necessary that Twoways relieves the patient’s pain (by giving Drug 1), Reasons NonNecessity implies that he has no reason to do so. So Weak Sufficiency does not imply that he has reason to give Drug 2. However, Reasons Non-Necessity seems implausible. Knowing that I will, or even that I am fated to, refrain from committing suicide in the next minute, for example, doesn’t undermine my judgment that I have reason to refrain. In any event, there are counterexamples to Strong and Ought Sufficiency, if not Weak Sufficiency, compatible with Reasons Non-Necessity. Suppose that Drug 1 fails in one out of every N cases, whereas Drug 2 works in every case. Even if it is necessary that Twoways gives Drug 1, it is not necessary that he relieves the pain. So, compatibly with Reasons Non-Necessity, there is reason to relieve the pain. Nevertheless, only weak reason transmits to the sufficient means of giving Drug 2, because in N–1 cases, giving Drug 2 is superfluous. By making the transmitted reason suitably weak (by the right choice of N), we can construct counterexamples to Strong and Ought Sufficiency.33