Suppose a pilot comes to us with a request for advice: “See, we’re at war with a villainous country called Bad, and my superiors have ordered me to drop some bombs at Placetown in Bad. Now there’s a munitions factory at Placetown, but there’s a children hospital there too. Is it permissible for me to drop the bombs?” And suppose we make the following reply: “Well, it all depends on what your intentions would be in dropping the bombs. If you would be intending to destroy the munitions factory and thereby win the war, merely foreseeing, though not intending, the deaths of the children, then yes, you may drop the bombs. On the other hand, if you would be intending to destroy the children and thereby terrorize the Bads and win the war, merely foreseeing, though not intending, the destruction of the munitions factory, then no, you may not drop the bombs.” What a queer performance this would be! Can anyone really think that the pilot should decide whether he may drop the bombs by looking inward for the intention with which he would be dropping them if he dropped them?
Thomson observes in the passage above that in deliberating about what to do, our attention is turned outwards rather than inwards. This is correct, but it's not clear why this should be trouble for the friends of intention. In deliberating about what to do, our attention is focused outward rather than inward upon the facts as we take them to be rather than our taking the facts to be a certain way. Many of us would still want to leave it as an open question as to whether the permissibility of an action depends upon the facts and not at all upon our beliefs about the facts. This is particularly true in epistemology, where there is nearly universal agreement that the permissibility of belief depends (in part) upon relations among beliefs and not just relations among the facts the agent happens to have in mind. Theoretical deliberation might be concerned with the question whether p is true, but it hardly follows that the normative standing of beliefs formed in response to that question depends entirely upon the external matters one focuses one's attention on in trying to settle that question. Why should things be different with practical deliberation?
One reason to think that our attention will be directed outwards rather than inwards when deliberating about what to do is that practical deliberation is concerned with what to do. While acting for some intention rather than another is something that happens and happens just when we do something, it is not itself something done. To see this, remember that the agent who decides to V could potentially V from any number of intentions. If she were obliged to V from one intention as opposed to another and this was something she did, it too could be the sort of thing that could be done from one intention as opposed to another. Again, if this is a doing, to be done from one intention rather than another, the agent would have to select between possible intentions. A vicious regress looms. It would seem that doing something from one intention rather than another would require completing an endless series of prior acts, something we cannot do. So, since doing something from one intention rather than another is not something we do, it is not something that we concern ourselves with in practical deliberation.
But doesn't this show that we have to reject the DDE since the DDE tells us that moral status depends, in part, upon intention? If acting for some particular intention rather than another is not something that we do, it is not something we can do, and so not something we can be obliged to do. This is correct, but this misses two things of potential significance. The first is that in the examples we are looking at, it is easy to follow the prohibition against V-ing from such and such an intention--do not V. So, there is no conflict between DDE and the principle that states that “ought” implies “can”. Second, it is not clear whether DDE is concerned in the first instance with acts or actions. Or, if you prefer, it is unclear whether the DDE is concerned in the first instance with types or tokens.
Thomson's point has bite, in part, because it would be absurd for the agent who realizes that the situation is the sort of situation in which someone may or must V in light of various considerations having to do with the nature of the situation to then ask whether she may V in light of the psychological factors that would explain her V-ing. If the question the DDE asks us is whether an agent could act from a suitable intention given the nature of the choice situation she faces, the reason Thomson's inward looking question is one we should not ask is that the question does not address something that the DDE is concerned with. If the DDE is concerned with act-types in the first instance, it tells us that the permissibility of an action will depend upon whether the agent could have acted for a sufficient reason, not whether her reasons for acting were sufficient.
Should we say that the DDE is concerned with acts rather than actions? I think we should. Moral principles like the DDE are in the business of telling us what our obligations are. What we are obliged to do is an act, not an action. When the reasons that apply to you demand that you do something (or refrain from doing something), the reasons are not concerned with the concrete details that distinguish two actions if they are of each of the same act-types. Also, as others have noted, the opposite view has trouble with negative obligations and the prospective nature of obligation. If you fulfill one of your negative obligations by not performing an act of a type, the omission does not seem to be a good candidate for an event and so we do not have a good candidate action to be the action that fulfilled the obligation. If you have an obligation you leave unfulfilled, there is something that is an obligation and we again have no good candidate action to be that obligation. [Some of these points are taken from Wedgwood and Schroeder from a discussion over at PEA Soup.]
Acting from such and such an intention or for such and such a reason is not something done, so it is not a type of act. Since what we are obliged to do is perform an act of a certain type, not any token action, and the DDE is concerned with our obligations, it is concerned with acts, not actions and the particular psychological antecedents that explain them. The question to ask, then, is not whether the agent's intentions in acting were such that they involved an intention to harm someone who had a claim against being so harmed but whether the agent could have acted as she did without such an intention. Thus, the fact that we do not look inward in deliberating about what to do is not a reason to think that intention has no bearing on permissibility.