Of course there is more pain, in one sense of pain. But one might reject the (natural) picture on which this automatically translate into greater harm (stating the example in terms of utility is thus question-begging).

I agree with you here. Note that, in my post, I refrain from describing the experience with the greater total amount of pain as the experience which has a greater total disutility. I also avoid describing the most intense pain as that which has the greatest disutility. (Kahneman and Kamm do not exercise similar restraint in the passages I quote from their work. And you’re right to call them on this.)

Supposed you tied well-being to preferences, not to hedonic states. On such a view, a brute preference to more pain distributed in this or that way would be sufficient to make that distribution better than the alternative.

But even if we tie well-being to preferences rather than hedonic states, there’s still the question of which preferences? Suppose that someone has already undergone one long trial and one short trial. Now you are subjecting him to a third trial. As he reaches the one-minute mark, you ask him whether he would prefer a complete cessation of the painful experience (which would amount to a short trial) or a continuation of it for an additional 30 seconds, but at a less intense level of pain (which would amount to a long trial). Suppose he expresses a preference for cessation. (As I maintained in my post, I think he would have to be either a masochist or crazy to have the contrary preference.) Suppose, however, that before subjecting him to this third trial, you had instead asked him which of the two trials that had already occurred he would prefer to have repeated. Perhaps he would have responded in the manner Kahneman predicts – i.e., he would prefer a repetition of the long trial. Surely we shouldn’t privilege the preference for the longer experience that is based entirely on memories of past experiences over the preference for the shorter experience that is expressed contemporaneously with the third trial. We should instead discount the entirely-memory-based preference on the grounds that our memories of experiences have been shown to be notoriously unreliable. (Kahneman, for example, has shown that we are very unreliable in recalling how long past experiences have lasted.)

I think it is a bias, but I also think you may be a bit too quick to reach this conclusion. We can reach it only if add certain (plausible) assumptions. In particular, we need a certain view of pain’s badness.

In cases where people prefer a decline to a rise, they do not deny that the total is worse. They just claim there is a further normative factor that tips the balance (as inqeuality might tip the balance in some interpersonal distribution of utility). But must we accept that the total is worse? Of course there is more pain, in one sense of pain. But one might reject the (natural) picture on which this automatically translate into greater harm (stating the example in terms of utility is thus question-begging). Here is one way (not the only one) in which one could reject this further step. Supposed you tied well-being to preferences, not to hedonic states. On such a view, a brute preference to more pain distributed in this or that way would be sufficient to make that distribution better than the alternative.

Perhaps it is an excess of philosopher’s caution, but I hesitate to say that Jones harms Smith in LT2 [by taking Smith’s medicine that he needs to consume in order to prevent the loss of his limb]. Jones certainly denies to Smith the medicine that would otherwise prevent the threat to Smith from running its course.

Suppose that Smith’s medicine is an antibiotic that is already in the process of being administered to Smith via a drip and is already starting to fight off the bacteria in Smith’s limb. Smith, however, needs a full dose of the antibiotic, otherwise he will lose his limb. Jones removes the yet-to-be administered portion of the dose from the bag that supplies the drip. Would you hesitate to say that Jones harms Smith in this case? If you would not, how do you distinguish this case from your LT2? If you would hesitate, how do you distinguish this case from the following case that surely involves harming: Jones turns off a respirator that Smith owns and that is keeping Smith alive during a temporary coma from which he will fully recover as long as he is oxygenated via the respirator for the period of the coma.

A separate remark:

I think that the fact that Jones steals from Smith in your LT2, thereby non-accidentally making him worse off than he otherwise would have been, is sufficient to ground the claim that Jones harms Smith in LT2. If you’d hesitate to endorse this sufficient condition of harming someone, I’d be interested to hear why.

Boat 1: In order to get five dying individuals whom you are transporting in your boat to the hospital in time to receive life-saving assistance, it is intuitively permissible for you to refrain from stopping en route to rescue a healthy individual who is stranded on a rock in a rising tide and who will eventually drown once the tide submerges the rock.

Boat 2: It is intuitively impermissible, by contrast, for you to run over someone on a raft even if that is necessary if you are to get five dying individuals whom you are transporting in your boat to the hospital in time to receive life-saving assistance.

Here it appears that the harming/not-aiding distinction explains the difference in intuition, rather than either the loss/no-gain distinction or the temporal or agential factors you mention.

Kamm suggests a comparison to show that the difference between the absence of benefit and loss is not morally significant in the Car cases:

Car Case 1: One is able to save the lives of five threatened with death in the back of one’s car by driving on to the hospital, and failing to aid a single person suffering from an existing, life-threatening medical condition on the side of the road.

Car Case 1’: One is able to save the lives of five threatened with death in the back of one’s car by driving on to the hospital, and failing to prevent a wild animal from killing an otherwise healthy person on the side of the road.

Kamm and Mike both judge driving on in these cases to be on a moral par. I agree.

Driving on in both cases is, with respect to the one, an omission of action. We might compare no-benefit/loss cases that involve an agent’s acting, too:

Strength Chemical Case 1: Left on his own, Smith will continue to grow in strength beyond normal strength to superhuman strength. Jones sprays Smith with a chemical that stops Smith’s growth in strength immediately and permanently, at normal levels.

Strength Chemical Case 2: Smith is superhumanly strong. Jones sprays Smith with a chemical that immediately reduces Smith’s strength to normal levels permanently.

In the first case, Jones prevents Smith from benefiting from superhuman strength. In the second, Jones deprives Smith of superhuman strength. I judge these cases also to be on a moral par. (We must be careful to bracket common, but inessential, differences between cases of this kind: the greater certainty of plans based the use of existing powers, as opposed to expected powers, etc.)

In defense of the moral parity of these cases, I suggest that whatever Smith might permissibly do to prevent Jones’ spraying in Case 2 he may do in Case 1. For example, Smith might permissibly knock Jones unconscious with a club to prevent the harm in Case 2, even if that would result in some permanent injury to Jones. He may also do the same in Case 1.

(As regards Kahneman and Tversky, would these cases elicit different responses from philosophically naïve audiences? About that I make no prediction.)

To turn to my own views on withdrawal-type cases, Neil’s case augmented by Mike’s comments yields the following:

Limb Threat 1: Smith will lose a limb unless he consumes some rare medicine. The medicine rightfully belongs to Jones. Jones removes the medicine from Smith and Smith loses the limb.

Limb Threat 2: Smith will lose a limb unless he consumes some rare medicine. The medicine rightfully belongs to Smith. Jones removes the medicine from Smith and Smith loses the limb.

Mike wants to say that, in the first case, Jones allows harm to Smith, and, in the second case, that Jones harms Smith. This difference is made by Jones having, or not having, a right to the medicine in the two cases. In this, Mike follows not only Kagan in the passage he cites, but Jeff McMahan and Kamm herself. (McMahan, Ethics of Killing (New York: Oxford, 2002), p. 379; “Killing, Letting Die, and Withdrawing Aid”, Ethics 103 (1993), pp. 250-79; “A Challenge to Common Sense Morality”, Ethics 108 (1998), pp. 394-418; Kamm, Morality, Mortality Vol. II (New York: Oxford, 1996), p. 28.)

I am less sure about the metaphysics of the situation. Perhaps it is an excess of philosopher’s caution, but I hesitate to say that Jones harms Smith in LT2. Jones certainly denies to Smith the medicine that would otherwise prevent the threat to Smith from running its course. I am more comfortable saying that Jones allows harm in LT1, but those who balked at this description because of Jones’ action would not meet with great resistance from me.

What I am much more confident about is that LT1 is on a moral par with what I call “standard allowings of harm”. A standard allowing of harm is a case in which an agent fails to act to aid, or prevent harm to, another, e.g., standing by while someone drowns. So, LT1 is on a par with a case in which Jones fails to send his rare medicine to Smith and Smith loses a limb. And LT2 is on a moral par with what I call “standard harmings”. A standard doing of harm, or a standard harming, is a case in which an agent creates a threat that harms another, e.g., when an agent attacks another with a club. So, LT2 is on a par with a case in which Jones poisons Smith causing Smith to lose a limb.

To support these claims, I would point out that Smith might permissibly use very serious force, perhaps even deadly force, to stop Jones from poisoning him resulting in a loss of a limb. So too might Smith permissibly resist Jones’ attempts to deprive him of his medicine in LT2. However, Smith may not use force, much less lethal force, in LT1, or in its standard allowing-of-harm analogue, to acquire Jones’ medicine.

The behavior in the Limb Threat cases is an example of what I call “denials of resources”, cases in which an agent deprives another of, or makes unavailable to another, a useful resource. In the article Thom mentions, only those resources that might prevent a loss from an existing threat are considered. The analysis also applies, though, to those cases in which a resource might produce a benefit.

The claim is that denials of resources are on a moral par with comparable standard allowings of harm when (a) the denying agent has a right to the resource denied and (b) the denied party has no right. Limb Threat 1 is a case in point. Denials are on a moral par with standard harmings when (a) the denying agent has no right to the resource denied and (b) the denied party has a right. Limb Threat 2 exemplifies this situation.

A right to a resource plays a much more important role in the moral status of denial cases than it does in either standard harmings or standard allowings of harm. The morally important metaphysical distinctions are thus between standard doings of harm and denials of resources, and standard allowings of harm and denials of resources. This is true whether denials of resources are doings of harm, allowings of harm, or neither. So, whether denials are doings, allowings, or neither is a much less important question for purposes of moral explanation than it has commonly been taken to be.

The idea about the independence in the wild animal case turns on how we individuate situations. My thought was this: there is a large mass of empirical evidence on how people individuate what is internal and what external to a case (in the moral case, the evidence has been gathered by Hauser). This evidence suggests that when causal processes begin is relevant to whether they are considered part of the same situation, and that whether the process is begun by an agent (that is, an animal or other self-animating being) is also relevant. If the causal process begins after the person has left, then it is independent of her; if it is begun by another agent, it is likewise independent. Kamm cannot ignore this evidence, since it is likely the same principles of situation individuation are at work in pumping her intuitions.

I am not sure that her counterexample compares like with like. In the wild animal case, the animal is independent of the agent’s actions. So the case might at best force a revision of the loss-versus-foregone-gains principles, according to which we put greater weight on loss than foregone gains, if the causal factors leading to each are held constant. That seems an irrational principle, and one that Kamm hasn’t shown isn’t at work in generating our intuitions.

I think, however, that, insofar as independence of the agent’s actions is concerned, Car Case 1 and Kamm’s wild animal variant of that case are like cases. In each case, that which would cause the death of the one person on the side of the road is something other than the agent’s actions. In Car Case 1, it’s the (unspecified) life-threatening medical condition he’s suffering. In the wild animal case, it’s the wild animal. We might, however, want to try to make these cases more similar by stipulating that, in Car Case 1, the one on the side of the road is in need of lifesaving aid because he’s already been attacked by a wild animal. We could add that, in all three cases (Car Case 1, Car Case 2, and Kamm’s variant of Car Case 1), the five are in need of life-saving assistance because they’ve been attacked by a wild animal.

But what of (c)? Plausibly, if I act to prevent the one person from regaining her limb I harm her (perhaps I deliberately move the medication she needs out of her reach). But my intuition is that doing so here is permissible. If it is permissible to harm here, this may be because doing so results in a forgone benefit, rather than a loss.

I would deny that the permissibility of harming in this case would be explained by the fact that you fail to provide her with a benefit rather than imposing a loss. I think that whether you harm her, and whether you act impermissibly, both instead depend on whether you have an independent right (e.g., a property right) to do that which prevents her from regaining her limb. If, for example, the medicine that you snatch away from her before she’s able to grab it is her medicine, then you’ve harmed her and you’ve done something impermissible. If, however, it’s your medicine for which she’s reaching, and you grab it first in order to deliver to the five, then you don’t harm her, nor do you act impermissibly. (Here I believe I’m tracking claims that Timothy Hall made in his PhD dissertation. I also believe that Hall was himself inspired to make these claims by remarks of Shelly Kagan on pp. 106-111 of The Limits of Morality.)

I am less confident than you are that her reply to the framing effect succeeds. Two points. First, I am not sure that her counterexample compares like with like. In the wild animal case, the animal is independent of the agent’s actions. So the case might at best force a revision of the loss-versus-foregone-gains principles, according to which we put greater weight on loss than foregone gains, if the causal factors leading to each are held constant. That seems an irrational principle, and one that Kamm hasn’t shown isn’t at work in generating our intuitions.

Second, consider this set of cases:

(a) I can save one limb of each of five people by cutting off the limb of one other person;
(b) I can choose to save the limbs of five people, but only by ignoring the plight of one other who without my aid will lose a limb;
(c) I can save the limbs of five people by acting so as to prevent one person from regaining a limb already lost.

Clearly (a) is impermissible while (b) is permissible or perhaps obligatory, just as both principles predict. But what of (c)? Plausibly, if I act to prevent the one person from regaining her limb I harm her (perhaps I deliberately move the medication she needs out of her reach). But my intuition is that doing so here is permissible. If it is permissible to harm here, this may be because doing so results in a forgone benefit, rather than a loss. It appears that sometimes losses-outweigh-foregone gains trump harming-worse-than-not- aiding. If that’s the case, then some of our moral intuitions are the product of an allegedly irrational principle, just as Horowitz claimed.