I think the distinction between the foreseen and the intended is central to a lot of our moral thinking, and that some form of the Principle of Double Effect is correct.
Here is a pair of cases that I find particularly difficult, however. This post owes things to a discussion I've been having with Daniel Hill.
Case 1: Matilda knows that a house contains two people, one an innocent and the other a terrorist. Matilda is flying over the house and can drop a bomb on it, and if she does so, both people will die. However, the terrorist will then be unable to detonate a bomb that would kill hundreds. Matilda has no other way of preventing the detonation of the bomb. Is it permissible for her to drop the bomb?
Case 2: This time Matilda is on the ground, and has a gun. The two residents of the house are in front of her. One of them is the innocent and the other is the terrorist. She can't tell which is which. The only way to prevent the detonation of the bomb is by killing the terrorist. (Why is wounding not good enough? Maybe from her position, she can only aim for the head, and so she'll either kill or miss.) Is it permissible for her to shoot both?
The first case seems to be just like the standard double effect case of tactical bombing where if you drop the bomb on the enemy HQ, innocent visitors to the HQ (e.g., spouses of officers), will also die. It is hard to distinguish Case 2 from Case 1, since it doesn't seem like it should morally matter whether one drops one bomb or takes two shots. But the action in Case 2 seems wrong if you have strong deontological intuitions. It sure seems like you're intentionally killing two people, where you know that one of them (but you do not know which--and that may change things) is an innocent.
So the challenge for the defender of double effect reasoning is to either show in a morally compelling way how Case 2 differs from Case 1, or show that the intuitions that the shooting in Case 2 is wrong are mistaken.
I will try for the first, but I don't know how morally compelling my story will be. I think it will only be compelling to those who find double effect reasoning compelling. Still I hope the story will have some plausibility. Let the two people in the house be Susan and Tricia. Matilda's intention in Case 1 is that the terrorist in the house die. By what means? By means of the place where the terrorist is being seared by an explosion. Matilda need have no intention in Case 1 regarding the non-terrorist, or regarding Susan qua Susan or Tricia qua Tricia. Her intention is explicitly about the terrorist as such.
Now consider Case 2. Suppose Matilda has Susan in her gunsights and squeezes the trigger. What are Matilda's reasons for so doing? The most plausible account seems to be something like this: "Susan may be a terrorist, and if so, then many lives will be saved by her death, so I will shoot her." In other words, the plan of action seems to be: "Shoot Susan dead, so that if she is the terrorist, the terrorist is dead." If that's the plan of action, then Matilda is (literally) aiming to kill Susan. And by the same token, Matilda is aiming to kill Tricia. Therefore, Matilda is intending the death of two persons, one of whom she knows to be an innocent. She knows, thus, that in her overall action plan there is an innocent whose death she is aiming at. And that is wrong.
Elsewhere, I have speculated that there are some actions that are only permissible with certain intentions. For instance, perhaps it is only permissible to assert with the intention of avoiding the assertion of a falsehood and perhaps sexual relations are only permissible with the intention of uniting maritally. It now seems quite plausible to me that intentional killing is like that. To kill someone intentionally and permissibly it is not enough that one believe that the person is an aggressor (or is probably an aggressor?) that one is duly authorized to kill, or however exactly the exceptions on the prohibition of killing should be put. The soldier or police officer needs to kill because the person is an aggressor that one is duly authorized to kill. The Allied soldier who justly kills a German soldier must do so because the German soldier is an aggressor. If the Allied soldier, instead, solely kills Helmut because Helmut is German or because Helmut has a long nose or because target practice is fun, the Allied soldier is morally corrupt. (What if the Allied soldier kills Helmut both because Helmut is an aggressor and because Helmut is German? I think more detail will be needed about the deliberative structure here, and I want to bracket this case.)
Now, let us suppose that in fact Susan is the terrorist. Then Matilda in intentionally killing Susan is not killing Susan because Susan is a terrorist. Rather, Matilda is killing Susan because Susan might be a terrorist. And that is not good enough. The intention to kill someone because she is a terrorist is compatible with love of that person, since doing justice to someone is compatible with love, and sometimes required by love. But that is not Matilda's intention.
This has an interesting implication for military ethics. It is often said to be necessary for soldiers to dehumanize their enemy in order to kill, to see them as enemies rather than as people, and this is often seen as a criticism of the military enterprise. But if I am right, it is morally required that the soldier kill Helmut under a description that includes something like "enemy aggressor" rather than simply under the description "Helmut" or "that man over there, who no doubt has a family who are awaiting his return." Perhaps in the ideal the humanity of the enemy, and the fact that he has a family who are awaiting his return, does enter into how the action is done--with compassion, sadness and only as necessary for due defense of the innocent. But Helmut's aggressor status needs to be in the soldier's intentions.
But let us go back to Case 2. One might cleverly object that it need not in fact be Matilda's intention that Susan die (Daniel Hill queried me about such an idea). It could perhaps be Matilda's intention that Susan die if she is a terrorist. Now, it is certainly possible to have such intentions. If one has, or thinks one has, a magic bullet that kills only terrorists, one could shoot Susan intending that she die if she is a terrorist. In that case, one's means to the conditional end that Susan die if she is a terrorist is shooting a bullet that discriminates between terrorists and non-terrorists. But in the actual Case 2, one brings it about that Susan dies if she is a terrorist by bringing it about that Susan dies: the conditional end is brought about, in this case, by the unconditional means. So one still intends that Susan die, as a means to the conditional end that Susan die if she is a terrorist.
But what if Matilda is a clever double effect casuist, and says: "My intention is that a bullet should go through such and such a location in space, and that if there be the head of a terrorist in that location, that terrorist should die"? However, I think this is an incorrect statement of Susan's intentions. Intentions aren't inner speeches. They embody our actual reasons for acting. Matilda's reason for sending the bullet to that location in space is that she can see Susan's head there. Her plan for making sure that the terrorist in that location should die seems to be that Susan should die, and hence if the terrorist is there, the terrorist should die. Susan's death still seems to be a means to the death of the terrorist in that location, if there be one there. And likewise for Tricia's death. I am not completely happy about this story, but it has some plausibility.
In Case 1, however, the aim is less personal, and that does actually matter: the only death aimed at is the death of "the terrorist", under that definite description. Certainly, we would expect Matilda to be much more traumatized by Case 2 than by Case 1 (and if she weren't, we would think there is something wrong with her), and we should take such trauma to be defeasible evidence for a morally relevant difference between the two cases.