By In Discussions, JMP Discussions, Normative Ethics Comments (32)

Journal of Moral Philosophy Discussion at PEA Soup: Molly Gardner’s “On the Strength of the Reason Against Harming,” with a critical précis by Fiona Woollard

Welcome to what will surely be an incredibly interesting and productive discussion of Molly Gardner‘s excellent paper “On the Strength of the Reason Against Harming.” This paper was published in the first issue of this year’s Journal of Moral Philosophy. They have kindly provided free access to the paper, which can be viewed or downloaded here. Fiona Woollard has written a critical précis, which is included below. Please join today’s discussion!


Fiona Woollard’s critical précis:

Molly Gardner’s paper is an interesting and insightful contribution to the debate on when we count as harming someone in a morally relevant sense.

Molly is interested in what the strength of reasons against harming can tell us about which account of harming is correct. A lot of arguments claim that a given account is better than its rivals because it is better at correctly distinguishing cases involving harm from cases where no harm is done.  As Molly points out, our explanations of moral judgments often appeal not just to whether harm was done but also to the strength of the reason against harming.  Sometimes it will be permissible to perform an action, even though it is harmful, because we only have a relatively weak reason to avoid the harm in question. This is an important and underexplored aspect of the debate.

Molly argues that, somewhat surprisingly, our judgments about the relative strength of reason against harming lend support to effective-relative accounts of harming.

Effect Relative Accounts versus Action Relative Accounts

Molly distinguishes:

Action Relative Accounts:  an action harms a person if and only if it makes her worse off in some respect than she would have been had the action not been performed.

Effect-relative Accounts:  an action may harm an individual in virtue of its effects on that individual, regardless of whether the individual would have been better off in the absence of the action.

The idea that I harm a person if and only if I make him worse off is appealing.  Nonetheless we might be tempted to endorse an Effect-Relative Account of Harming because Action-Relative Accounts clash with our intuitions about three types of cases:

Selecting for Poor Health. Barbara wishes for her child to have poor health. She uses in vitro fertilization and screens the embryos for genes that cause poor health. Finding such an embryo, she implants it, and the fertility clinic discards the other embryos. The implanted embryo grows into a child named Billy. Billy develops poor health, and as a consequence of having poor health, Billy experiences hardship, pain, and suffering. Nevertheless, Billy’s life is on the whole worth living. (Gardner, p. 76)

Shooting Match: Through no fault of his own, Victor has made two terrible enemies, Adam and Barney, who have both sworn deadly vengeance upon him. Barney is just about to shoot and kill Victor. Barney is protected by a bullet-proof, sound-proof shield so that Adam can neither stop him forcibly nor dissuade him. Adam knows this, but Victor’s death by another’s hand will not satisfy his thirst for vengeance. Adam shoots Victor and Victor dies from the bullet wound (Woollard 2013, p. 8, quoted on Gardner, p. 76)

Bigglesworth’s Revenge: Dr. Evil has locked Jones inside a small, steel box that also contains a Geiger counter, a small amount of a radioactive substance, a small hammer, and a flask of poison gas. Everything is rigged so that there is a 50 percent chance that an atom of the substance will decay at time t. If an atom of the substance decays at t, the discharge of the Geiger counter will cause the hammer to smash into the flask of poison gas, and Jones will be killed. If an atom of the substance does not decay at t, then Jones will be released from the box. Irritated that the setup is too complicated, Dr. Evil’s assistant, Bigglesworth, sneaks over to the box with a baseball bat and swings the bat against the side of the box. The force of the swing causes the hammer to smash into the flask just before time t, and Jones dies (Gardner, p. 77).

Selecting for Poor Health is a non-identity case (without the action, the victim would not exist); Shooting Match is a pre-emption case (if the agent did not perform the harmful action, someone else would have harmed the victim in the same way); Bigglesworth’s Revenge is an indeterminacy case (there is no fact of the matter about how well off the victim would have been without the harmful action).  In each case, intuitively, the victim has been harmed, but would not have been worse off without the harmful action.

The Strength of Reasons against Harming

Molly argues that, surprisingly, our judgments about the strength of reasons against harming support an effect-relative account of harming.  This is surprising because it might at first glance seem like action-relative accounts do better here.  Action-relative accounts have a natural explanation of why harmful actions are wrong: they are wrong because they make others worse off than they would otherwise be.  This leads to a natural explanation of how strong our reasons against harming are: the greater the decrease in well-being of the victim, the stronger the reason against harming.

Molly also discusses my own argument about the strength of reasons against harming. I’m interested in the Non-Identity Problem in cases like climate change: intuitively, it is wrong for us to pollute because we will harm future generations, but, because our policies about pollution affect who comes to be born, we don’t make any individual worse off by polluting.  I argue that when we harm someone but do not make them worse off, our reasons against harming are weaker than if we harm someone and make them worse off.  We can’t fully solve the Non-Identity Problem by arguing that we can harm someone without making them worse off: this still leaves our reasons not to pollute much weaker than we think they are.

Suppose that by shooting Victor, Adam could save the life of a third party, Sarah.  It is permissible for Adam to shoot Victor to save Sarah (given that Victor was going to die anyway).  It would not be permissible for Adam to save Sarah’s life by shooting another person who was not going to die anyway.

Molly argues that Shooting Match does not show that our reasons not to harm are weaker when we do not make the victim worse off.  Shooting Match involves ‘redundant harming’: the action causes a harm that the victim would have suffered anyway. Molly argues that the reason against harming is weaker if the harming is redundant.  Consider the following cases:

Inducing Paralysis: It is Leland’s 5th birthday, but unfortunately, a piece of debris, A, is flying through the air towards Leland’s midsection. If Bernard does nothing, A will hit Leland and paralyze him from the waist down. Bernard’s only other option is to push Leland into the path of another piece of debris, B, which would cause Leland a qualitatively identical injury; however, pushing Leland into the path of B would prevent Sarah from having a broken arm. Bernard pushes Leland into the path of B.

Selecting for Paralysis. Enid is choosing which embryo to implant into her uterus. Embryo B will become an individual named Sheldon. B has a mutation that will cause Sheldon to become instantly and permanently paralyzed from the waist down on his 5th birthday. However, implanting embryo B will prevent Sarah from having a broken arm. Enid implants B.

Neither Bernard nor Enid make their victim’s worse off than they would have been without the action.  Inducing Paralysis involves redundant harming.  Selecting for Paralysis does not. Leyland would have been paralysed whatever Bernard did. Sheldon would not have been paralysed if Enid had acted differently.  Molly argues that Enid’s behaviour is morally worse than Bernard’s – and deduces that our reasons against redundant harming are weaker than our reasons against non-redundant harming.  Because most non-identity cases involve non-redundant harming, my example does not show that our reasons against harming in non-identity cases are weaker than we previously thought.  An effect-relative account of harming can explain our intuitions about the strength of reasons not to harm by endorsing the following principle:

(R) Other things being equal, the reason against redundantly harming an individual is weaker than the reason against non-redundantly harming an individual.

There is a worrying whiff of circularity about the appeal to intuitions about a non-identity case to rescue our intuitions about non-identity cases.  I am also not convinced that redundancy that explains the weakness of the reasons against harming Leyland.  We can put together versions of Inducing Paralysis which do not involve redundant harming: perhaps debris B would kill Leyland or cause Leyland to lose his sight.  When we vary the harms, the key question for working out if Bernard’s behavior is permissible seems to be whether the alternative harm is more or less severe than the one Bernard would inflict i.e. whether Bernard’s action leaves Leyland worse off.   This supports my suggestion that our reasons against harming are much weaker if we do not thereby make the victim worse off.

2 More Principles About the Strength of Reasons not to Harm

Molly proposes two more principles that an effect-relative account of harming should endorse: the Degree of Harm Principle and the Inevitability/Avoidability Principle. She argues that an effect-relative account of harming that endorses these three principles offers the best explanation of our judgments about the relative strength of reasons against harming.

Degree of a harm (def.): The degree of a harm is directly proportional to the amount by which an individual’s level of well-being on some dimen­sion would be higher if the individual existed and the state of affairs that is the harm had not obtained.

(D) Other things being equal, the reason against harming is stronger, the greater the degree of a harm.

By appealing to this principle, an effect-relative account can explain why the reasons against selecting an embryo with paralysis are stronger than the reasons against selecting an embryo with mild eczema.  Action-relative accounts cannot explain these differences because in neither case do we make the victim worse off than they would have been without the harmful action, so we do not count as harming in either case.

I agree that an effect-relative account of harming which endorses the Degree of Harm Principle is in a better position to account for some of our intuitions about the relative strength of reasons against harming than an action-relative account.   Nonetheless, I still hold that our reasons not to harm are stronger when we make the victim worse off.  We should not endorse either a pure effect-relative account or a pure action-relative account, but recognise different morally relevant ways in which we can harm someone, adding a fourth principle to the three suggested by Molly:

Action-relative principle:  Other things being equal, the reason against harming is stronger, the more the harmful action makes the victim worse off than they would otherwise be.

Love this idea? Nominate it for the Annual PEA Soup Awards!

32 Responses to Journal of Moral Philosophy Discussion at PEA Soup: Molly Gardner’s “On the Strength of the Reason Against Harming,” with a critical précis by Fiona Woollard

  1. Hi Molly,

    I really enjoyed your paper and it helped me more clearly see what can be said in favor of an effect-relative account of harm. At the same time, I wonder whether action-relative accounts can avoid a few of the problems that you address in the beginning.

    One of the three supposed problems for action-relative accounts is that it cannot hold that, in Shooting Match, Adam wrongs Victor in virtue of harming him unjustly. That is true. But couldn’t the proponent of an action-relative account can hold that Adam is blameworthy for acting with malicious intent? He could presumably also be blameworthy for his disposition to harm Victor unjustly since it seems he would have shot Victor even if Victor’s death were not overdetermined. Since the action-relative accounts can accommodate these judgments, I don’t see the intuitive advantage to holding that Adam wrongs Victor in virtue of harming him unjustly. Although, this could just be the result of my own theory-laden intuitions.

    A second supposed problem for action-relative accounts is that it cannot hold that, in Biggleworth’s Revenge, Bigglesworth wronged Jones in virtue of harming Jones since there is no fact of the matter about whether Bigglesworth’s action made Jones worse off than he otherwise would have been. I agree that the action-relative account is committed to this, but I don’t necessarily see that as a problem. The reason why is that action-relative account can say that Bigglesworth violated a subjective obligation by smashing the bat against the side of the box. In doing so, Bigglesworth failed to maximize expected value and so acted wrongly. Since blameworthiness tracks subjective obligations, it seems to me that proponents of an action-relative account of harm capture the idea that Bigglesworth acted wrongly and is blameworthy, which I take to be the correct verdict in this case.

  2. I would like to thank Molly for her interesting and insightful paper, and Fiona for her precis and response. I have enjoyed working through Molly’s paper, which has helped my clarify and further develop my own views on harm. This discussion came at the perfect time for me, as I am in the process of revising a couple of chapters on harm for a book I am finishing on scalar morality. It will come as no surprise to Molly that I don’t share her views on harm. My own approach is a contextualist version of the counterfactual comparative view. I have far too much written to include in one blog post, so I will get the ball rolling with some comments on the first two of Molly’s cases, explaining why I don’t think they support the effect-relative account:
    There are at least two crucial distinctions, which are ignored, in building a case for the supposed weakness of action-relative accounts. Consider the non-identity case. I don’t even have a slight inclination to say that the child is wronged in that case. In her initial discussion of the case, Molly says “Billy seems to have been wronged by Barbara’s action”. Tellingly, however, in a later reference to the case, Molly says that Barbara’s action in this case “seems wrong”. I agree that Barbara’s action in this case seems wrong (on the assumption that the category of wrongness has a place in moral theorizing, either as a basic category, or as a contextually constructed one). But, crucially, this does not entail that Billy, or anyone else, is wronged by the action. One of Parfit’s insights, in introducing the non-identity “problem” to the literature, is that an action can be wrong, without it wronging anyone. It is only if we uncritically accept the principle that a wrong action must wrong someone or other that we are tempted to conclude that Barbara’s action wrongs Billy. And then, if we do, erroneously, claim that Billy is wronged, we look for an account of what the supposed wronging consists in. That Billy is harmed by Barbara’s action is then put forward as the most likely explanation. Once we discard the dogma that a wrong action must wrong someone, the “problem” is solved. The reason why Barbara’s action seems wrong is that Barbara had what were intuitively better options available to her (implanting different embryos, without birth defects).
    Molly claims that effect-relative accounts can also cope with preemption cases better than action-relative accounts. But, again, this requires that we agree that Adam (in her example) wrongs Victor, when he shoots and kills him, even though Barney would have killed Victor, had Adam not killed him, and there was nothing Adam could do to prevent Barney killing Victor (except to kill Victor himself, of course). But it’s not at all clear that we should agree that Adam wrongs Victor. We, of course, have a strong negative reaction to Adam’s action, but that is because we are told that he has a thirst for vengeance, and really wants to kill Victor himself. This brings up the second distinction that is frequently ignored (or misunderstood) in these sorts of discussions. We can judge the moral quality of actions, and we can judge the moral quality of an agent’s character. Frequently, these judgments go hand in hand, but sometimes they don’t. Adam is clearly not a great guy. Our strong intuitive reaction against the example may well be justified by the immorality of his character, as displayed in this action, rather than the immorality of his action. To see this more clearly, imagine that Adam doesn’t, in fact, want Victor dead, but realizes that he can’t prevent Barney killing Victor. Furthermore, imagine that Adam knows he can produce some good effect only by being the one to kill Victor himself. Suppose that he can prevent an innocent child, Agnes, from suffering a painful broken leg, which will permanently disable her. Suppose further that Adam knows that, if he does kill Victor in order to save Agnes from a broken leg, there is a good chance that he will be punished for it, because the law, as we all know, is an ass. Now, if Adam kills Victor, it seems clear that he is doing nothing even a little bit wrong, but rather is committing a heroic act of self-sacrifice to save an innocent child considerable pain and lifelong disability. It’s not enough to say that the harm being caused to Victor is redundant, and so the reason against doing it is less strong than in another case, so that, all things considered, Adam’s act is justified. There is simply no reason at all against Adam killing Victor in this case. Only by ignoring the crucial distinction between judgments of actions and judgments of character are we tempted to say otherwise. This, and a further point about the use of intuitions, also explains why, contrary to Fiona Woollard, there isn’t even a weak reason against non-comparative harming. Molly reports that Woollard claims that it would be wrong for Adam to shoot Victor, if the shooting would save Sarah from a scraped knee. I agree that the intuition that Adam would behave wrongly in this case is not unusual. However, there are two important considerations that show that such intuitions are not, in fact, good evidence that it really would be wrong for Adam to shoot Victor in this case. The first is the relevance of character, that I have just explained. Human character is not infinitely malleable. We want people to have pretty strong dispositions not to shoot innocent people. Probably, to be strong enough, such dispositions would make it really hard for us shoot an innocent person, even if we thought that a little bit of extra good, such as saving someone the pain of a scraped knee, could come of the shooting. So, once again, at least some of the strength of our negative reaction to this version of the example is explained as a judgment on Adam’s character. Second, it is by no means clear that our intuitions actually react to the cases as described. It is common for hypothetical cases in the ethics literature to include very strong stipulations about what is possible in the case. In this case, it is stipulated that it is certain that Barney will shoot, if Adam doesn’t, that Adam cannot possibly prevent Barney from shooting, and that it is certain that Barney’s shooting Victor will kill him. Now consider the real world. That is, the world that actually shapes our intuitions, which ethicists then demand we bring to bear on worlds that are increasingly removed from it. In the real world, how justified would you be in believing that, if you didn’t shoot Victor, Barney would be certain to do so himself, that there is absolutely no way for you to stop Barney shooting Victor (except by shooting Victor yourself), and that Barney’s shot is absolutely certain to kill Victor? You wouldn’t, of course, be justified in believing any of these things. In the real world, then, there would be some chance that Victor would survive, if you didn’t shoot him, and less chance that he would survive, if you did. The difference you make to the chances of Victor dying by shooting him might be quite small, but, given the seriousness of death, it is still big enough that a benefit such as preventing a scraped knee would probably not be big enough to justify taking the shot. Once we move to benefits like preventing broken arms, the situation may well be different. I know that the example includes the stipulation that Barney is certain to shoot and kill Victor, if Adam doesn’t. However, we can stipulate until we’re blue in the face. No such stipulations can guarantee that the intuitions we have about the case are actually about this highly unrealistic case, rather than a version of it, which is closer to what might actually occur, or that such intuitions should be given much weight.

  3. I have one other worry. You formulate principles (D), (V), and (R) in terms of the strength of reasons against harming. However, I presume that they also apply to the strength of reasons to prevent others from harming. Either way, I worry that they will not only be insufficient to account for intuitive judgments in many cases, but also generate the wrong verdicts in many cases. Here is one such case.*

    Preventing a Shooting: Two people are about to be shot at the same moment by two different aspiring assassins. You have time to prevent just one person from being shot by shouting the aspiring assassin’s name, which will distract them, causing them to miss their shot. You can prevent PERSON 1 from being shot by shouting “JOHN, DON’T SHOOT!”. You can prevent PERSON 2 from being shot by shouting “WILKES, DON’T SHOOT!”.

    If PERSON 1 is shot, he will survive, but suffer 1,000 units of physical pain unique to being shot. Yet, if PERSON 1 is not shot, another aspiring assassin will immediately proceed to hit PERSON 1 over the head with a baseball bat. Doing so will cause PERSON 1 to suffer 1,000 units of physical pain unique to being bludgeoned with a bat. By stipulation, PERSON 1’s total well-being will be the same if he is shot and maimed as if he is hit with a baseball bat. Finally, John is not particularly invested in shooting anyone and can easily be persuaded to forgo the attempted assassination.

    If PERSON 2 is shot, she will only suffer 500 units of physical pain unique to being shot. If PERSON 2 is not shot, she will go on to live a long and happy life. By stipulation, her total well-being will be much higher if she is not shot than if she is shot. Now, Wilkes is determined to assassinate PERSON 2 and cannot be persuaded to forgo attempting the assassination.

    Here is my worry. It seems to me clearly true that you have much more reason to prevent PERSON 2, rather than PERSON 1, from being shot. However, my understanding is that (D), (V), and (R) entail that you have more reason to prevent PERSON 1 from being shot. The degree of harm for PERSON 1 is double what it is for PERSON 2, so (D) favors saving PERSON 1. It is not the case that PERSON 2 being shot is less inevitable than PERSON 1. If anything, PERSON 1 being shot is less inevitable, so (V) either doesn’t favor either action or favors saving PERSON 1. Finally, both harms are non-redundant harms, so (R) cannot favor saving PERSON 1 or PERSON 2.

    Unless I have misunderstood (D), (V), and (R), it looks like they, coupled with an effect-relative account of harm, are going to generate incorrect verdicts in cases in which (i) there is an action-relative harm that affects a person, yet (ii) that person doesn’t suffer the greatest degree of harm, so (D) favors harming someone unaffected by the action-relative harm and (iii) (V) and (R) don’t favor any action.

    One could potentially avoid this problem by positing the fourth principle that Fiona suggested. But this raises questions about how to weigh the principles against one another when they conflict. To get the intuitively correct verdicts, I worry that this fourth principle would have to be lexically prior to the others. But then, why not just get rid of the other three principles and adopt an action-relative account of harm? Admittedly, these other principles could preserve intuitive judgments in non-identity cases. But I’m happy to bite the bullet (if it really is a bullet) in non-identity cases and go with the more parsimonious action-relative account.

    * Although this case concerns the strength of your reasons to prevent others from harming, a structurally identical case could be given that concerns the strength of your reasons against harming. To amend the case accordingly, simply imagine that you are forced to shoot either PERSON 1 or PERSON 2.

  4. Jens Johansson says:

    Hi Molly,

    Thanks for a very stimulating paper. According to the version of the effect-based account that you propose on p. 11, an action harms S just in case it causes a harmful state for S, where a harmful state for S is a state such that living with it is better for S than living without it would be. But consider this case (taken from your Ergo paper):

    Dim Vision: Jones has been blind for many years as a result of a retinal damage. Recently, Dr. Smith has developed a surgical operation that can repair some but not all of the damage. Dr. Smith operates on Jones and improves his vision from a state of blindness to a state in which Jones can see, but not very well: Jones now has what we will call *dim vision*.

    Suppose that (contrary to what you assume in that paper) if Jones existed without dim vision, then Jones would have had better vision, and would as a result have been better off. This could be, for example, because Dr. Smith would still have operated on Jones, and another doctor — who actually remains passive — would have done some additional work on Jones’s eyes. It may be that the state of dim vision is harmful for Jones here, but it seems clear that Dr. Smith’s action does not harm Jones in any way. (We can assume that there was no alternative act available to Dr. Smith which would have made things better for Jones.) But that is what your proposed account implies, as Dr. Smith’s action causes Jones to have dim vision.

    Furthermore, your three principles about the severity of harm (pp. 11-14) yield that there is a moral reason against Dr. Smith’s action. That, too, seems wrong to me.

  5. I also some worries about what Molly says about the difference between non-comparative non-redundant harming and non-comarative redundant harming. In trying to argue for the claim that, at least sometimes, the reason against non-comparative non-redundant harming will be stronger than the case against non-comparative redundant harming, Molly compares the cases of inducing paralysis (Bernard) and selecting for paralysis (Enid). She declares that her judgment is that “Although Enid’s action may be permissible, it is still a morally worse action than Bernard’s”. This, she says, suggests that other things being equal, an action that qualifies as non-comparative harming but not redundant harming is morally worse than an action that qualifies as non-comparative redundant harming. But this doesn’t seem right. Suppose we accept that Enid’s action is worse than Bernard’s. What is the difference between the two cases that explains this difference? Molly seems to be assuming, as evidenced by her use of the phrase “other things being equal” that the only morally relevant difference between the cases is that one involves non-comparative non-redundant harming and the other involves non-comparative redundant harming. But this ignores the most glaringly obvious difference between the cases. Bernard does not have available to him an option with overall better results. Enid does. Enid could, presumably, choose an embryo that wouldn’t suffer paralysis. In the bizarre world of this example, such a choice wouldn’t cure her broken arm, unlike the choice of implanting Sheldon. But, clearly, a world in which a child suffers permanent paralysis on his fifth birthday, but the mother is spared a broken arm, is worse than one in which there is a different child who doesn’t suffer paralysis, but whose mother does suffer a broken arm.
    Even though Molly argues that non-comparative non-redundant harming is, other things being equal, worse than non-comparative redundant harming, she does at least imply that there is at least some reason opposing even non-comparative redundant harming. This suggests that, in a choice between two options that only differ in that one involves some non-comparative redundant harming, and the other doesn’t involve any kind of harming, there is at least some reason to prefer the latter. But this is implausible. Consider the following example: Two Hour Headache. Your child has an unpleasant but non life-threatening condition, that afflicts her with frequent two hour headaches. The best available treatment for this condition is a pill that has the effect of taking away the second hour of the headache. After using the pill for a while with success, you discover that there are in fact 2 versions of the pill: pill A takes an hour to start working, and then removes the second hour of headache. Pill B works immediately by preempting the first hour of headache, replacing it with a qualitatively identical hour of headache, then removes the second hour. From the perspective of the patient, the pills work identically. The only difference between giving pill A and giving pill B is that the latter involves non-comparative redundant harm. Pill B causes a one-hour pain, that is qualitatively identical to one the patient would suffer anyway. Pill A doesn’t do this. They both remove the second hour of pain, that the patient would have otherwise suffered. Suppose that you discover that the bottle of pills you bought to treat your child is a mixture of pill A and pill B. Would you have any reason at all to try to figure out which ones were which? Of course not. As far as the morality of your treatment of your child is concerned, the pills are identical. But one of them causes some non-comparative redundant harm, and the other doesn’t. But this provides no reason at all, not even a smidgen, to prefer the other pill. It’s not enough to say that the reason against non-comparative redundant harming is smaller than that against other types of harm. It’s simply non-existent here. The reason is simple. It makes no difference to the welfare of the patient. What this demonstrates, of course, is that the notion of making a difference is central to anything that is morally interesting or relevant about harm. When we have certain intuitions, or make certain judgments, at least the justified ones, about acts harming patients, we are picking out certain relevant comparisons. There is no simple algorithmic method for identifying the relevant comparisons in all cases. This is where context comes in.

  6. Nathan Hanna says:

    Hi, Molly – Great paper! Glad to see it being discussed here. I’m curious about something you say on p83. You mention a specific effect-relative account that says that a subject is harmed if they’re caused to be in a harmful state. You then say this:

    “Such an account can explain in a straightforward way why there is a pro tanto reason against harming. Namely, to harm someone is to cause a state of affairs that detracts from her well-being. What makes it true that the harmful state of affairs detracts from the individual’s well-being is that, if the individual existed and the state of affairs did not obtain, the individual would be better off in some respect.”

    I’m worried that this entails that certain intuitively harmful states of affairs and acts aren’t harmful. Suppose you push me out of the path of a car, causing me to sprain my ankle. If I’d been hit by the car, I’d have been paralyzed. Intuitively, you harm me pro tanto (but not overall), since you cause a state of affairs that’s intuitively harmful to me: I’ve got a sprained ankle. Your claims don’t seem to get this result, though. They don’t seem to entail that you harm me in any sense because if I hadn’t sprained my ankle, I’d have been paralyzed and been worse off.

    Maybe I’ve misunderstood the quoted passage, though. Maybe the thought is this: a state of affairs is harmful for a subject if she would have been better off had it not obtained *and everything else stayed the same*. That entails that my spraining my ankle is harmful and that you harm me by causing it. But this take on the passage seems to entail that certain intuitively harmless states of affairs are harmful. Suppose a grandfather gives lavish birthday presents to his young grandchildren. But this year, he decides that his eldest granddaughter is too old to be spoiled like that. So he just sends her a nice card. Intuitively, her not getting a present from him this year isn’t harmful to her (assume, e.g., that her feelings aren’t hurt) and he doesn’t harm her by not send her a present. He just fails to benefit her. But my reinterpretation of your passage seems to entail that the state of affairs is harmful and that he harms her by causing it. Had that state of affairs not obtained then, keeping everything else the same, she would have gotten a lavish present.

    So I’m worried that the view expressed in your passage either counts some intuitively harmful events as harmless or that it counts some intuitively harmless events as harmful. Maybe I’ve misunderstood, though. Or maybe my intuitions are off. I’d be interested to hear what you think!

  7. Duncan Purves says:

    Hi Molly! I love the paper. Not surprisingly, like some other folks in this thread, I want to resist getting rid of the counterfactual comparative element in determining the harmfulness of events. I want to say a couple of things in defense of the action-relative account in reply to the preemption case:

    You say that the preemption case counts against the action-relative account of harm because it seems that Adam wrongs Victor and that the action-relative account cannot explain why. But this is not a problem for the action-relative account, because the account is merely an account of harming, not of wronging. Maybe Adam violates Victor’s rights without harming him, because whether I violate your rights does not depend solely on whether I harm you (e.g., the harmless trespass).

    But I’m more interested in hearing more about how (and the extent to which) the effect-relative account proposed on p. 11 avoids the preemption problem. The effect-relative says harming is causing a harmful state for some individual, and that a harmful state for an individual is one such that living with that state is worse for the individual than living without that state would be. Describing what makes a state a harmful state, you say, “what makes it true that the harmful state of affairs detracts from the individual’s well-being is that, if the individual existed and the state of affairs did not obtain, the individual would be better off in some respect.”

    So Adam harms Victor on this account because he causes Victor to die from the bullet wound, and Victor is worse off than he would have been had he existed and not died from the bullet wound. That is, dying from a bullet wound is a harmful state of affairs, and Adam harms Victor because he causes this state of affairs. But my first worry is how to individuate the relevant state of affairs. Is the state of affairs Victor’s *actual* death from the bullet wound? Victor is not worse off than he would have been had he existed and that state of affairs not obtained. If some other state of affairs, distinct from Victor’s actual death by the bullet wound is the harmful one, it is not immediately clear how Adam causes it.

    One other aspect of the effect-relative account seems strange to me. It entails that the state of affairs *Adam’s causing Victor’s dying of a bullet wound* is not a harmful state of affairs for Victor, because Victor would not have been better off in any respect if he existed and the state of affairs did not obtain. I think this is a counterintuitive result on its own, but it is especially odd since the state of affairs merely consists of (by the lights of the effect-relative account) the harming event and the harmful state of affairs in virtue of which Victor counts as being harmed.

  8. Hi, Everyone.

    Wow! Thank you for these great and thoughtful comments. Thanks especially to Travis and PEA Soup for hosting this discussion, and to Fiona for writing a great precis.

    I’m going to respond to each point in a separate comment, so please stay tuned. –Molly

  9. Hi Nathan and Duncan, good comments. It seems we share worries about the effect-relative (and/or non-comparative, to the extent that they are different) approaches to harm. As you quote Molly saying, “What makes it true that the harmful state of affairs detracts from the individual’s well-being is that, if the individual existed and the state of affairs did not obtain, the individual would be better off in some respect.” You’ve both pointed out that an account such as this can generate some puzzling results in individual cases. I agree, but I also think that there’s a deeper problem with this kind of approach, that goes beyond simply generating counterintuitive results (which can be of limited value in constructing ethical theories). In defining harm in relation to these kinds of counterfactuals, the account becomes potentially divorced from reality, and thus from anything that should guide our actions in situations of choice. For example, in non-identity cases, it is simply (nomologically) impossible for the child to exist without the relevant state of affairs (some form of disability, usually) obtaining. Why, then, are we considering what the world would be like in any respect under conditions that we have admitted are beyond our reach? That would be like criticizing my choice of taking the bus to work, on the grounds that it is more time-consuming and less energy-efficient than my using the plant-powered teleportation device that you have just read about in the latest sci-fi novel. Sure, if such devices existed, and I had one, it would be better to use that to get to work. But since they don’t, consideration of what the world would be like if they did is completely irrelevant to the ethics of my choice. This doesn’t actually show that such considerations aren’t relevant to what people are talking about when they use the term “harm”. But it does show that, if and/or when such considerations are part of what people are talking about with terms like “harm”, what they are talking about has precisely no actual moral relevance. Duncan’s comment also raises an important theoretical point, that I have made in several different papers. Any definition of something in terms of what it would be like if that thing didn’t obtain (even with such specifications as that the person exists anyway) suffers from the problem of fixing the alternative. If we are wondering what life would be like for Victor, if he existed, but without his actual impairment, we need to know precisely what his physical condition would be instead. This is where context comes in. It is simply unavoidable.

  10. On Individuating Harms (a reply to Alastair, Travis, Nathan, and Duncan):

    I think various assumptions about how to individuate harms might be behind the following worries:

    •Alastair’s worry about the pill that removes 1 hour of a headache and replaces it with 1 hour of a different headache
    •Travis’s worry about the bullet that causes 1,000 units of pain but prevents 1,000 units of a different pain
    •Nathan’s worry about the lifesaving action that causes a broken ankle
    •Duncan’s worry about Victor’s actual death not being worse than the preempted death
    •Jens’s worry about Dim Vision

    Also, Jeff McMahan didn’t weigh in on this blog, but he does spend a lot of time worrying about how to individuate harms in his book The Ethic of Killing.

    I didn’t argue for a specific account of harming in this paper (my goal was to defend a *type* of account of harming, rather than a specific, substantive account), but I do formulate a more precise, substantive account of harming in my Ergo paper (“A Harm-Based Solution to the Non-Identity Problem.”) So to respond to the above list of worries, I’m going to draw a bit on the account of harming that I offer in the Ergo paper.

    On that account, a sufficient condition for harming someone is that you cause them a harm. Notice that you don’t have to cause them all of the harms—just causing someone *a* harm is sufficient. So even if there are some states of affairs that are not harms and you cause them, that doesn’t mean that you didn’t harm the person. And even if there are some states of affairs that *are* harms and you don’t cause them, that doesn’t mean that you didn’t harm the person. Causing just one harm is enough!

    In response to Duncan’s worry, then, I’m going to point to a harm that Adam caused Victor: the harm is dying when he did. Dying when he did is a harm for Victor because if he hadn’t died when he did, he would have been better off in some respect. Also, Adam caused Victor to die when he did. It doesn’t matter, in this case, that there are other non-harms that Adam caused.

    In response to Jens’s worry, I want to make sure I understand the case. Jens, are you suggesting that by operating on Jones, Dr. Smith prevented the passive doctor from improving Jones’s eyesight even more? If so, then here’s a harm for Jones: not having better eyesight. Now, did Dr. Smith cause this state of affairs? I say in my Ergo paper that omissions (which I call “negative states of affairs”) can be effects, but there I’m mainly supposing that the causes of these omissions are acts of procreation. I’ve been thinking more about Batman and the golf clubs lately, and now I’m not sure whether I think that preventions also count as causes of omissions. (In other words, suppose A is about to benefit C, but B steps in and prevents A from benefiting C. Does B cause C to be without the benefit?) But if they do, then Dr. Smith harms Jones by causing Jones to not have better eyesight. Dr. Smith also benefits Jones by causing Jones to not be completely blind. So there would be both a harming and a benefiting in this case, which I think sounds right, actually. On the other hand, if preventions are not causes of omissions, then Dr. Smith only benefits Jones.

    What I just said about Dim Vision, I think, is also relevant to one of Nathan’s concern about the grandfather, if we think that his failure to give the presents counts as a prevention. (Maybe there was some thought in his mind like, “Oh, wait! She’s too old for a present now!” that prevented his kindlier nature from shining through.) If the grandfather’s change of mind caused the kid to be without a present, then it harmed her. I suppose I would bite the bullet on that one. (But if there was no change of mind at all—if there was just a plain and simple failure to give a present—then it didn’t harm her.)

    AAACK!!!! More comments keep coming in!!! I’m freaking out because I haven’t read Alastair’s newest comment yet. I’m going to post this now and then add more later. Please stay tuned…

  11. More on Individuating Harms:

    Let me now address these worries:

    • Alastair’s worry about the pill that removes 1 hour of a headache and replaces it with 1 hour of a different headache
    • Nathan’s worry about the lifesaving action that causes a sprained ankle

    First, to address Nathan’s worry about the sprained ankle, I note in my Ergo paper that I want to talk about being better off in some respect, rather than better off, full stop. So in the sprained ankle case, the person is better off with respect to having a longer life but worse off with respect to her ankle pain.

    Let me now address Alastair’s worry about the pill. Alastair writes, “The only difference between giving pill A and giving pill B is that the latter involves non-comparative redundant harm.” First, a small point: I don’t like the term “non-comparative harm” because my own account of harm is a comparative one, even though it is effect-relative. So it would be better to say “The only difference between giving pill A and giving pill B is that the latter involves redundant harm.” But even though that would be *better*, it’s still not true. There is another difference between giving pill A and giving pill B, namely, that by giving your kid pill B, you give her an extra benefit that doesn’t come along with pill A. The extra benefit is the taking away of the initial hour of headache. (That is to say, administering pill B harms the kid by causing her a first hour or headache, and it benefits her by removing a different first hour of headache, and it benefits her again by removing the second hour of headache. Giving pill A only benefits the kid by removing the second hour of headache.) So we would need another case pair where we really are holding all else equal.

  12. I’m about to be busy for most of the rest of the day with work (curse the need for a paycheck!), but I’ll add one more post for now. Suppose that we want to agree with Molly’s judgments about what acts harm whom in her three examples. Can an act-relative account give us these results? Yes. The clue is in the role that context plays in ordinary judgments about harm. Here’s a contextualist counterfactual account of harm: An act A harms a victim V just in case A results in V being worse off than she would have been had the appropriate alternative act been performed instead. “Appropriate” here is an indexical, whose referent is fixed by the context of utterance (cf. “Today is a good day to die”). In the preemption case, the conversational context could select as the appropriate alternative a world in which neither Adam nor Barney shoots. In the non-identity case, the conversational context could select as the appropriate alternative a world in which Billy actually exists, but without health defects. In the indeterminism case, the conversational context could select as an appropriate alternative a world in which Bigglesworth doesn’t smash the box, and the risk of death stays at 50 (on the plausible suggestion that raising the probability of death is itself harmful). Whether the conversational context would actually select those comparisons as the referents of “appropriate alternative” is a psychological question. In the preemption case, the world in which neither Adam nor Barney shoots is not so very different from the actual world that it couldn’t be the one we fix on. Recall the difficulty we have getting students to take seriously the conditions of various trolley examples (“but why don’t you simply run down the track and pull the one guy off it?”). The fact that, in the non-identity case, the facts of biology seem to tell us that Billy couldn’t have existed without health defects certainly doesn’t prevent us considering such a world as the appropriate comparison. At least Molly shouldn’t think so, given what she says about the sorts of comparisons that can ground judgments of inevitability/avoidability. The great advantage of the counterfactual contextualist account is that it can both explain certain judgments, and show how they are morally irrelevant. In footnote 8, Molly points out that we can compare “how the victim would fare had the effects of the action obtained to how she would fare had she existed and the effects of the action not obtained.” This is clearly supposed to be relevant to judgments in non-identity cases. Perhaps some people really do think that selecting for disability harms the resultant child (who nonetheless has a life worth living), because they are comparing the actual life of the child with the life they suppose that very child would have had had she existed without the disability. If the judgment that the act of selecting for this child makes her worse off than she would have been had she existed without this disability is the one being expressed by the harm judgment (or perhaps in some way is supposed to ground the harm judgment), then we can see both how someone might be saying something true (or at least understandable) by such a claim, and that the judgment itself is completely morally irrelevant. Simple reflection on the fact that it is not, in fact, possible for the child to exist without the disabilities tells us that such comparisons, no matter how linguistically appropriate they may be, cannot be morally relevant, in the sense that they cannot provide reasons suitable for guiding actions. As philosophers, as opposed to politicians, we have to accept that reality is relevant to morality. The set of beliefs commonly held in a world in which people believe that single women who live with cats are malevolent witches, who use supernatural powers to cause suffering to innocent people, can explain why people in that world could claim that someone who lives with their cat but no humans can deserve to suffer. Such claims depend crucially on beliefs we know to be false (that women who live with cats and no humans have supernatural powers that they use to inflict suffering on innocent people). Likewise, given what we know about human reproduction, the claim that not only is Billy harmed by being selected for, but that what grounds the harm claim is itself morally relevant to the morality of the choice can be seen to be simply mistaken. The world in which Billy exists, but without the disabilities, is inaccessible from this. Given that I can’t bring about that world, we can see that any comparison between that world and this is not relevant to the morality of my choice. The worlds that I can bring about, in which people other than Billy exist without disabilities, are relevant to my choice. That I can, but fail to, bring about one of them explains why my choice to bring about the Billy world is morally criticizable, even though it doesn’t wrong Billy.

  13. Let me know address Travis’s worry about preventing harm and Fiona’s suggestion that I add a 4th principle.

    In response to Travis, I think the case is very similar to Alastair’s pill case. So I’d say the same thing: if John shoots PERSON 1, he benefits him by preventing 1,000 units of baseball bat pain, but he harms him by causing 1,000 units of bullet pain. The benefit is pretty significant, and that is a feature that can be weighed against the pro tanto strength of the reason against harming, so that the all-things-considered reason against John’s shooting PERSON 1 is weaker than the all-things-considered reason against Wilkes’s shooting PERSON 2.

    I am opposed to adding the action-relative principle that Fiona suggests. I think you can get essentially the result that Fiona wants by taking the degree of harm principle more seriously.

    However, I’m happy to add this principle:

    The Benefit Principle: The all-things-considered reason against harming someone is weaker if and to the extent that the harmful action also benefits the victim.

  14. Teresa Bruno says:

    Hi Molly, Great paper! I enjoyed reading it and I’m very sympathetic to your view. I would like to suggest a case that I find particularly compelling in favor of your view. It would be great to hear what you think about it. Suppose that Oscar has been unfairly imprisoned. We could fill in the details in different ways, but say that he was misidentified by a witness and sentenced to five years in prison. Oscar led a rather dull and meaningless life before being incarcerated. During his time in jail he works at the library and learns a great deal about the horrors of mass incarceration. While in jail and especially after, he becomes a tireless advocate for the rights of those unfairly imprisoned. His life now is meaningful and he is much better off than he would be otherwise.
    The action relative account of harm give us the wrong result that Oscar wasn’t harmed by being incarcerated unfairly because he isn’t worse off than he would be otherwise. Your view in contrast can account for this case because Oscar is all things considered better off by being incarcerated, but it is worse off in one respect, namely that his autonomy was violated because he wasn’t allowed to shape his own destiny. Presumably, exercising our autonomy makes us better off, so Oscar is harmed because being incarcerated “detracts from the individual’s well-being” in some respect.
    An advantage of this case is that we can prompt intuitions about it without taking us too far from the actual world. This might help with Alastair Norcross’s worry about Shooting Match that our intuitions might not be reacting to the cases as described, but to a version of those cases that is closer to the actual world. I think it might also help to answer worries about whether we are reacting to other features of the case. In Shooting match, we might be reacting to the vicious character of the agent, for instance. I think that in my case it is right to point out that part of the wrongness comes from the fact that Oscar was incarcerated unfairly, but it is plausible that we are also reacting to the fact that Oscar was deprived of his freedom and therefore harmed.

  15. Jens Johansson says:

    Hi Molly,

    Thanks for your response. I did not mean to suggest that Dr. Smith’s action prevented the passive doctor from improving Jones’s eyesight even more. By operating on Jones, Dr. Smith did his part of the improvement, and the passive doctor independently omitted to do his. It does not seem unrealistic to suppose that the details are such that the nearest world where Jones exists but does not have dim vision is, not one where Dr. Smith does not operate, but one in which the other doctor makes some further improvements. Whether or not dim vision is a harm for Jones, I cannot see that Dr. Smith’s action, which causes this state, harms him in any way.

    I won’t be able to write any more comments today after this one — but many thanks for an interesting discussion, both to you and to everyone else!

  16. Just popping in again during a 15 minute break between meetings. Regarding Molly’s point about my 2 hear headache case that giving pill B gives an extra benefit that pi;; A doesn’t give, namely taking away the first hour of headache (along with giving a qualitatively identical hour), this suggestion won’t help Molly’s cause. The reason is that either giving that extra benefit cancels out the harm of giving the hour’s headache, or it doesn’t. Suppose first that it doesn’t. Even though pill B gives the benefit of taking away the original hour’s headache and the harm of replacing it with a qualitatively identical one, the net effect is negative. That is, if that were all that giving pill B did, there would be a reason not to give pill B. If this is the case, it would matter whether I gave pill A or pill B to my child, but, of course, it doesn’t. That leaves the possibility that the benefit of taking away the hour’s headache and the harm of replacing it with a qualitatively identical one cancel each other out. This gives the result that pill A and pill B are morally identical, which is the right result. However, it also gives the result that the reason’s for and against Adam shooting Victor also cancel each other out. This is also the right result, but not the one Molly wants or defends. What is going on, of course, is that thinking of the effects of pill B in terms of the harm of causing the pain and the benefit of preventing the qualitatively identical pain is really just another way of thinking of the supposed harm here in terms of the counterfactual comparative account. Regarding Teresa’s example, the counterfactual comparative account can give the result she wants. Even though the unjust imprisonment benefits Oscar all things considered, which is exactly the right result, it clearly harms him in some respects. He has his freedom restricted for a while. Had he not been unjustly imprisoned, he wouldn’t have suffered that particular harm. It’s also true that the act of unjustly imprisoning Oscar couldn’t have been expected in advance to benefit him. Of all the confusions in the harm debate, the one between harm in some respect and harm all things considered is the most important one to clear up.

  17. Sorry for all the typos in my previous post. In a rush.

  18. Hi Molly,

    Thanks for your reply! There are structural similarities between my case and Alastair’s, so it probably isn’t surprising that the response I was going to give is similar to the one he already gave. If (D) favors saving PERSON 2 because the harm of a 1,000 units of bat pain cancels out the benefit of missing out on the 1,000 units of bullet pain, then it seems like (D) just is (or is very close to) the action-relative principle Fiona suggested. I’m worried that I’m missing something. Is there any non-non-identity cases where they come apart?

  19. Hi, Teresa.

    Thank you! And yes, exactly. I have another paper in which I argue that my account gets the right results in exactly the kind of case you describe. That paper is under review right now, but if you’d like to read it, just let me know where to email it. –Best, Molly

  20. OH CRAP. I SAID THE WRONG THING. You win this round, Alastair and Travis, but you haven’t won the war.

    Here’s what I should have said:

    First, let’s get clear about our headaches and our times. Let’s say that this whole thing is happening from 2 to 4 p.m. If you give your kid pill B, then the headache she has from 2 to 3 p.m. is called “Headache X” and the headache she would have had from 2 to 3 p.m. is called “Headache Y.”

    What I should have said is that Headache X is not a harm at all. Why? Well, the following is not true:

    COUNTERFACTUAL A: If she hadn’t had Headache X, then she would have been better off in some respect.

    COUNTERFACTUAL A is not true because if she hadn’t had Headache X, then she would have had Headache Y, and that’s no better in any respect. Therefore, giving your kid pill B is not a case of harming, and so it is not a case of redundant harming.

    On the other hand, Adam’s shooting Victor still IS a case of harming. This counterfactual is true:

    COUNTERFACTUAL B: If Victor hadn’t died when he did, he would have been better off in some respect.

    The difference between the two cases is that Adam causes the general state of affairs of Victor’s dying when he did, but giving your kid pill B does not cause the general state of affairs in which she has a headache (at all) from 2 to 3 p.m. It only causes the specific state of affairs in which she has Headache X. (And while having a headache–at all–is a harm, having Headache X, in particular, is not a harm.)

  21. Hi Molly, I’m not understanding your latest move. You say “Adam causes the general state of affairs of Victor’s dying when he did, but giving your kid pill B does not cause the general state of affairs in which she has a headache (at all) from 2 to 3 p.m. It only causes the specific state of affairs in which she has Headache X. (And while having a headache–at all–is a harm, having Headache X, in particular, is not a harm.)” But why not say that Adam doesn’t cause the general state of affairs of Victor’s dying when he did, but only the specific state of affairs of his dying death X, which is not a harm, because he would have died death Y at the hands of Barney, if Adam hadn’t killed him? I really don’t see how you can justify the move in one case but not the other. Is there supposed to be something special about death, as opposed to other kinds of harms? What if the Adam and Victor case involved, instead of death, severe pain and disability as a result of the gunshot? In this case, if Adam doesn’t shoot Victor and cause him severe pain and disability, Barney will shoot him instead and cause a qualitatively identical but numerically distinct pain and disability. Any intuitions about the original Adam and Victor case would apply to this case too.

  22. Hi Molly,

    Alastair beat me to the punch again. I initially thought that this was the type of move you were going to make, which is why the harms in my ‘Preventing a Shooting’ case were stipulated to be unique to the cause of the harm (i.e. shooting or bludgeoning). They are qualitatively distinct and, we can suppose, even occur at different times. The only morally significant common feature is that the first type of harm (being shot) is as bad, all things considered, for a person as the second type of harm (being bludgeoned).

    Given how different these harms are, I was thinking that my original ‘Preventing a Shooting’ case would have to be treated like the ‘Shooting Match’ case with Victor and Adam. That is, it would have to be treated as a case of harming. But if that’s right, then (D), (V), and (R) coupled with an effect-relative account of harm would (I think) generate the incorrect verdict in my ‘Preventing a Shooting’ case. Alternatively, the ‘Preventing a Shooting’ case and the ‘Shooting Match’ case could both be treated like Alastair’s ‘Pill’ case. That is, they could not be considered cases of harming. But making that move, it seems, requires relying on an action-relative harm principle.

  23. Fiona Woollard says:

    I’m a bit puzzled by Alastair’s stipulation that Headache X and Headache Y are distinct. My sense is that it only make sense to talk about two different headaches if the headaches are qualitatively distinct. If the headaches are qualitatively distinct, then I do see a difference between the two cases. The Pill which causes a different headache seems to me to be a kind of limiting case of the type of situation in which we harm someone in order to benefit them overall. And I think this harm is morally significant – for example, it is relevant for issues of consent.

  24. Hi, Travis, Fiona, and Alastair,

    This is super helpful. I’m still thinking about all this. I have to teach until 9 tonight, but I will try to do another post after that. Thank you for having this discussion with me. –Molly

  25. Hi, Everyone.

    Sorry to post this after such a delay. But I think I really needed to take a bit more time to think things through fully before responding. In my haste to respond on Monday while everyone was still interested, I ended up saying a few things I regret.

    If you’re just happening upon this blog post, feel free to ignore almost all of what I said in the comments above, except the parts where I explained what Headaches X and Y were. Otherwise, THIS comment (I hope!) is my definitive response to some of the really great objections people raised.

    First, the worries articulated by Travis, Fiona, and Alastair are interesting because they show exactly why the redundant harm principle (R) has such a restricted scope—that is, they show why (R) is to be used in cases only where we’re talking about the *same* harm, rather than (a) two different harms that would have had the *same* overall impact on someone’s well-being or (b) two different harms that would have had different impacts on someone’s well-being.

    Travis’s shooting case is one where we’re talking about two different harms that would have had the same overall impact on someone’s well-being. In that case, the feature that rules out the applicability of (R) is precisely the feature that allows us to give an alternative explanation for the weakness of the reason against John’s shooting PERSON 1. That feature is Travis’s stipulation that the harm of 1,000 units of bullet pain is different from the harm of 1,000 units of bat pain. Insofar as these are different respects in which PERSON 1 could be harmed, the degree of harm principle (D) implies that John has a stronger reason against shooting PERSON 1 than he would have had, say, if shooting PERSON 1 prevented exactly the same state of affairs that it caused. However, there is also a degree of benefit principle (B) that is analogous to (D). It says that

    (B) Other things being equal, the reason for benefiting is stronger, the greater the degree of a benefit.

    And insofar as preventing bat pain and the bullet pain are different respects in which PERSON 1 could be benefited, the degree of benefit principle (B) implies that John has a stronger reason for shooting PERSON 1 than he would have had, were the two benefits the same. The relatively stronger reason against shooting PERSON 1 can be balanced against the relatively stronger reason for shooting PERSON 1, and although harm-based reasons are generally stronger than benefit-based reasons, the former can still be attenuated by the latter; the upshot is that there’s only a weak all-things-considered reason against shooting PERSON 1. There’s a much stronger all-things-considered reason against shooting PERSON 2, and that’s the right result.

    I can say very similar things in the kinds of cases Fiona was worried about. Fiona wrote, “We can put together versions of Inducing Paralysis which do not involve redundant harming: perhaps debris B would kill Leland or cause Leland to lose his sight. When we vary the harms, the key question for working out if Bernard’s behavior is permissible seems to be whether the alternative harm is more or less severe than the one Bernard would inflict i.e. whether Bernard’s action leaves Leland worse off.” But what I want to say is slightly different. Compare these three cases:

    Case 1: If Bernard does nothing, debris A will hit Leland and paralyze him from the waist down. If he pushes Leland into debris B, then B will paralyze Leland from the waist down.

    Case 2: If Bernard does nothing, debris A will hit Leland and paralyze him from the waist down. If he pushes Leland into debris B*, then B* will kill Leland.
    Case 3: If Bernard does nothing, debris A will hit Leland and paralyze him from the waist down. If he pushes Leland into debris B’, then B’ will give Leland a very light bruise.

    Intuitively, in Case 1, there is a very weak reason against pushing Leland; in Case 2, there is a very strong reason against pushing Leland; and in Case 3, there is a strong reason that favors pushing Leland. My view gets all of these intuitively plausible results without invoking an action-relative principle; I need only invoke D and B.

    Alastair’s case is a different. In Alastair’s case, we can’t invoke B because Pill B does not benefit your kid by preempting the first hour of headache. It’s not true that if the kid had headache Y, she would have been worse off in some respect. There’s no respect in which it would have been worse for her to have headache Y rather than headache X. Thus, without (R), we’d get the result that there’s a pretty strong reason against giving your kid pill B, and that would be the wrong result. Luckily, (R) was formulated to deal with precisely this kind of case. (R) gets us the result that there is only a weak reason against giving your kid pill B.

    Alastair’s objection, of course, was that this was the wrong result. He says that there is no reason, not even a weak one. But what I should have done all along is bite the bullet. It’s not a very distasteful bullet to bite—we’re talking about the difference between a very weak reason versus no reason at all. Notice, as well, that the reason against giving your kid Pill B is even weaker than the reason Barney has against pushing Leland in Inducing Paralysis, and that reason is, itself, weaker than the reason Adam has against shooting Victor in Shooting Match. That’s because the degree of harm of having a headache from 2 to 3 p.m. is not as great as the degree of harm of being paralyzed from the waist down, and the degree of harm of being paralyzed from the waist down is not as great as the degree of harm of dying when Victor died.

    Let me recap: I said above that the worries articulated by Travis, Fiona, and Alastair are interesting because they show exactly why the redundant harm principle (R) has such a restricted scope—that is, they show why (R) is to be used in cases only where we’re talking about the *same* harm, rather than (a) two different harms that would have had the *same* overall impact on someone’s well-being or (b) two harms that would have had different impacts on someone’s well-being.

    Travis might worry that the more I account for these sorts of cases, the more my view starts to look like an action-relative view. After I posted one of my failing comments above, he wondered whether there were any non-non-identity cases where the two views come apart. Well, now that I have bitten the bullet about Alastair’s pill case, the views come apart there. And given what I said above, the views still come apart in Shooting Match, Inducing Paralysis, and Bigglesworth’s Revenge.

    Here’s an added bonus of my view, which I did not discuss in the paper. The added bonus is that action-relative views are usually bundled together with the concept of counterfactual dependence, whereas on my view you get a clear commitment to harming as a causal concept. There are all sorts of wonderful benefits that come when you characterize harming as a causal concept, rather than one that involves counterfactual dependence. If we appeal to the correct account of causation—one that recognizes a distinction between a cause and a mere condition—then we can say that the set of things your action causes, in some particular case, is much smaller than the set of things that depend counterfactually upon your action. For example, if Thomas Edison hadn’t invented the light bulb, then I wouldn’t have gotten a parking ticket last December. Given this true counterfactual, action-relative folks are going to have to say that Edison harmed me, which is implausible. But since Thomas Edison’s act of inventing the light bulb didn’t cause me to get the parking ticket, I don’t have to say he harmed me, and that’s good. (I’ve got another paper about this issue under review right now.)

    Ok, I’m going to post this, and then I’m going to post one more thing in response to Jens. Thank you again.

  26. Jens, thank you for clarifying your worry; I apologize for misunderstanding it the first time. Here’s the original case:

    Dim Vision. Jones has been blind for many years as a result of retinal damage. Recently, Dr. Smith has developed a surgical operation that can repair some but not all of the damage. Dr. Smith operates on Jones and improves his vision from a state of blindness to a state in which Jones can see, but not very well: Jones now has what we will call dim vision.

    (By the way, originally was Matthew Hanser’s case. I cited him in the footnote, but I don’t think my footnote made it clear that he was the one who originally thought of the case, so I’m sorry about that.)

    Jens, you wrote, “It does not seem unrealistic to suppose that the details are such that the nearest world where Jones exists but does not have dim vision is, not one where Dr. Smith does not operate, but one in which the other doctor makes some further improvements.”

    I think that in order to make this ordering of possible worlds come out true, we’re going to have to add a few details to the original case. Here’s the new case, with the extra details added:

    Dim Vision 2: Jones has been blind for many years as a result of retinal damage. Recently, Dr. Smith and Dr. Bloggs have developed a surgical operation that can perfectly repair all of the damage. The operation has two stages: In stage 1, Dr. Smith does something, and in stage 2, Dr. Bloggs does something else. Jones goes into the operating room, expecting to come out with perfect vision. However, during the operation, Dr. Bloggs fails to do his part. Consequently, Jones wakes up and can see, but not very well: Jones now has what we will call dim vision.

    Suppose Jones asks the question: “Why do I have dim vision?” If we’re dealing with the original Dim Vision case, then the answer to his question is, “Because Dr. Smith operated on you.” It is true, in other words, that in the original case, Dr. Smith caused the state of affairs in which Jones has Dim Vision. But I said in my paper that even though Dr. Smith caused this state, it was not a harm, for if Dim Vision had not obtained, Dr. Smith wouldn’t have operated, and Jones would have been completely blind—he would not have been better off in any respect. (I rely on backtracking counterfactuals a lot, which is one reason my view gets similar results to the action-relative account. But I have good reasons for relying on backtracking counterfactuals.)

    Now suppose we’re talking about Dim Vision 2, and Jones asks, “Why do I have dim vision?” In this case, we wouldn’t give the same answer. We wouldn’t say, “Because Dr. Smith operated on you.” Instead, we’d say, “You have dim vision because Dr. Bloggs failed to do his part of the operation.” Thus, Dr. Smith is no longer the cause of the state of affairs in which Jones has dim vision. It’s true that in this case, Dim Vision is a harm; if Jones had not had Dim Vision, then he would have had perfect vision, so he would have been better off in at least one respect. But Dr. Smith didn’t cause this harm; Dr. Bloggs did.

    My suggestion, then, is that if you change the details of the case so that dim vision goes from being a non-harm to being a harm, you also have to change some of the causal details as well. So you either have a case where dim vision is not a harm, although Dr. Smith caused it, or you have a case where dim vision is a harm, but Dr. Smith did not cause it.

  27. Hi Molly,

    Thank you for taking the time to write such a detailed and helpful follow-up reply. It’s interesting to see where your view comes apart from action-relative accounts. Even though it does, I am happy to see that the differences between the two views are more narrow than I initially thought.

    Thanks again for the great discussion!

  28. Fiona Woollard says:

    Thanks for the response Molly. I’m going to need to think a lot more to work out if you have converted me, but the view is certainly looking attractive.

    I love the bit about separating causation from counterfactual dependence. It has always bothered me that the action relative accounts push together the question of whether you have harmed and whether the victim has suffered a harm, making I.e. The doing / allowing distinction invisible.

  29. Thanks for the responses Molly. It has helped me see even more clearly why I disagree with pretty much everything about your approach to harm. In separating causation from counterfactual dependence in the view (presumably in concert with a non-counterfactual theory of causation), you have a view of harming which can’t explain its moral relevance. Maybe that’s fine. Ultimately, I think harm has no moral relevance in itself, but that’s because the notion is infected with contextualism, even to the extent of making comparisons with nomologically impossible situations potentially relevant (e.g., non-identity cases). What I do think is obvious is that our tendency to invest harm with moral significance comes from the cases where the relevant comparisons are with situations that we really could have brought about instead. The basic moral notion is that of making a difference. The only reason to move beyond that to a non-comparative account is a desire to get the results we want in thought experiments. But, even if that is a relevant theoretical consideration, it’s really easy to explain and account for those intuitions with a comparative account. And, in response to Fiona, making the doing/allowing distinction invisible is, of course, exactly the right thing. The tendency to invest the doing/allowing distinction with moral significance has led to untold suffering, and is pretty much entirely explained by the selfish desire to justify our own neglect of the interests of others.

  30. Fiona Woollard says:

    “And, in response to Fiona, making the doing/allowing distinction invisible is, of course, exactly the right thing. The tendency to invest the doing/allowing distinction with moral significance has led to untold suffering, and is pretty much entirely explained by the selfish desire to justify our own neglect of the interests of others.”

    Perhaps unsurprisingly, I disagree with you there Alastair. It’s worth noting that rejecting the doing/allowing distinction doesn’t automatically lead most people to follow a very demanding morality – I think it is much more likely that most people would accept a much less demanding theory and see themselves as permitted to do more harm than we currently accept. So I am dubious about our claim that investing this distinction with moral significance has led to untold suffering. I also think it can be defended as necessary for anything to genuinely belong to us even our own bodies – and that there are very good reasons that morality should recognise each person’s body as belonging to him or her. (See Woollard, Doing and Allowing Harm, for a full version of this argument.)

    But putting all that aside, even if we don’t think that the doing/allowing distinction has moral significance, there are clear problems with defining harming in a way that prevents us from discussing whether it matters whether we do harm or merely allow it. For one thing, we are going to create needless problems with intuitions – because intuitively “harming” involves actively doing harm. Even if you are going to argue that what ultimately matters is whether your behaviour makes a difference, I don’t think the right way to go about this is to analyse ‘harming’ in that way. Instead, we should try to find analyses that match our ordinary concepts more closely and *then* argue that these ordinary concepts fail to fully reflect what matters morally.

  31. Jens Johansson says:

    Hi Molly,

    Thanks for your reply on the dim vision case. I don’t agree but it is good to know how you are thinking about this. I will discuss this sort of case, and your view more generally, in a paper (work in progress) on harm-based approaches to the non-identity problem.

  32. Alastair, it’s okay that you disagree with my view, but you have to stop calling it a non-comparative account of harming! It is a comparative account of harming, even though it is effect relative. (Also, I do endorse a counterfactual account of causation, but the account I like best is Alex Broadbent’s reverse counterfactual account, which relies on backtracking counterfactuals instead of forward-tracking ones. He’s got a good argument about why this actually does capture the intuitive notion of “making a difference.”)