Let A1-A6 stand for six distinct agents. Let V1-V6 stand for six distinct potential victims. Let T1-T6 stand for six distinct and successive times. And let C1 stand for some set of agent-centered constraints. Those who endorse agent-centered constraints accept the following:
(1) It would be impermissible for an agent, A1, to violate a set of constraints, C1, even if this is the only way (short of doing something even worse) for A1 to prevent A2-A6 from each violating C1.
There are at least two possible rationales for (1):
The Agent-Focused Rationale (AFR): The explanation for (1) lies with the fact that agents ought to have a special concern for their own agency — that is, for what they themselves do as opposed to what they merely allow to happen.
The Victim-Focused Rationale (VFR): The explanation for (1) has nothing to do with the thought that agents should have a special concern for their own agency. Rather, as Frances Kamm has claimed, “the agent's own act is special only in that it makes him come up against the constraining right” of his would-be victim. This constraining right acts as a barrier against the permissibility of treating him in certain ways, such as treating him as a means to the minimization of rights violations overall.
Many deontologists endorse VFR. Frances Kamm and Richard Brook certainly do, and I believe that Eric Mack and Jerry Gaus do as well. In support of VFR, these philosophers often appeal to intuitions such as this one:
(2) It would be impermissible for A1 to kill a victim, V6, through the introduction of some new lethal threat even if this is the only way (short of doing something even worse) for A1 to prevent V1-V5 from being killed by the lethal chain of events that A1 earlier initiated.
To make things a bit less abstract, imagine that A1 earlier set up a bomb that will kill V1-V5 unless A1 now shoots V6 and places her body over the bomb. (I borrow this case from Kamm.) Assume that shooting V6 is the only way (short of doing something even worse) for A1 to save V1-V5. Assume, then, that A1 cannot save the five by throwing her own body on the bomb. On VFR, it is impermissible for A1 to kill V6 even though this is the only way for A1 to prevent herself from becoming the killer of V1-V5. As advocates of VFR would put it, V6 has a constraining right that constrains A1 from using V6 as a means to minimizing even her own violations of this constraining right.
But (2) doesn’t support VFR over AFR, for the advocate of AFR can also endorse (2). The advocate of AFR could, for instance, hold that it is not A1's killings that A1 should be most concerned to minimize; rather, what A1 should be most concerned to minimize is instances of her treating people as a mere means — e.g., instances of her intending to cause a person’s death as a means to achieving her own ends. And, by shooting V6, A1 wouldn’t minimize the instances in which she treats someone as a mere means. Indeed, nothing A1 can do now can undo the fact that she has already treated each of V1-V5 as a mere means. So treating V6 as a means to minimizing the deaths she causes only adds to the number of instances in which she treats someone as a mere means. So if the relevant set of constraints includes a constraint against treating people as a mere means, then the advocate of AFR can endorse (2).
To test whether it is AFR or VFR that offers the best explanation for (1), we should, then, consider, not whether we endorse (2), but whether we endorse:
(3) It would be impermissible for A1 to violate C1 at T1 even if this is the only way (short of doing something even worse) for A1 to prevent herself from violating C1 on each of five separate future occasions: T2-T6.
The fact that approximately 80% of the respondents thought that Smith should break his promise to Tom on Saturday so that he could then fulfill his promises to Rick and Harry on Sunday suggests that most of us reject (3) and accept AFR. Indeed, less than 10% of the respondents thought that Smith should keep his promise to Tom, which is what Smith should do if VFR is correct. So it seems that most of us accept AFR, not VFR. Of course, one might worry that none of the respondants are people who actually endorse an agent-centered constraint against breaking one's promises; one might worry, for instance, that PEA Soup readers are all act-utilitarians. But, as the other survey showed, this is not the case. A sizeable majority of the respondents thought that Smith should not break his promise to Tom in order to enable Jones to keep his promises to Dick and Harry. So most of the respondants do endorse an agent-centered constraint against breaking promises. These two polls, then, would seem to provide us with some evidence that AFR as opposed to VFR offers the best explanation for (1).