...If my picture of the difference between consequentialist and non-consequentialist forms of reasonings is right, the real difference between the two relies in judgments about the scope of the variables that should enter into “moral calculation”.
According to certain forms of act-consequentialism, (taken both as an agent theory and as an act theory) all good and bad facts ought to be included in this calculation. According to others, only foreseeable good and bad consequences ought to be included in this calculation.
A non-consequentialist theory of "what is right", will limit the range of facts that should be assessed in order to perform the right action, compared to both types of consequentialist theories (meaning both tipes of act-consequentialism), in a specific way.
The fact that certain good or bad events take place ought not enter into moral calculation. A non-consequentialist theory would argue that I cannot kill an innocent to prevent another one to be killed, (or that I cannot kill an innocent in order to prevent another person to kill an innocent in order to prevent another innocent to be killed.) A think what a deontological requirement should be seen as doing is ruling out from view considerations on what good my killing the innocent should achieve as irrelevant from practical reasoning. For another example, a non-consequentialist theory will deny me the right to kill an innocent now in order to avoid killing another one in the future. Facts about what I shall achieve by killing an innocente are excluded from the calculation.
This may be too strong a way to characterize what is for a view to be non-consequentialist. If all facts about what I shall achieve by killing a person are irrelevant, then the deontological prohibition says that I shall never kill an innocent even when doing so would prevent a massacre.
In order to avoid this result, we simply have to think deontological constraints as a sorts of partial filter, that gradually lets facts about consequences become relevant to our calculations, as we reach a certain threshold (it can be thought that, once a certain threshold is reached, it becomes morally wrong to ignore the consequences.)
According to this view, the difference between consequentialist and non-consequentialist moralities (moralities that include some deontological prohibitions) consist in the fact that non-consequentialist moralities limit the range of considerations that can count in determining whether one is acting rightly or wrongly.
This is coherent with what psychologists call “bounded rationality” approach to human reasoning, and with evolutional reconstruction of moral sense.
Psychological evidence suggests that human beings must use heuristics in order to act optimally given the practical limitations that human decision making is subject to, such as a time limit, and finite processing capacities.
A non-consequentialist ethics is the only kind of ethics that makes sense to agents , because it is appropriate for agents whose decision procedures are subject to empirical limitations, such as those of which theorist of bounded rationality talk about.
This account seems to provide no justification for the remorse that deontological or virtue theory views should attach to killing an innocent even when this is (supposedly) what must be done in order to prevent a massacre from occurring.
But I think that the justification that it provides is sufficiently good. We can describe this remorse as tracking down the loss in moral integrity, that is, as a loss in one's confidence about being able to live up to the standards morality requires. As I said a deontological view of morality is THE moral view an agent should have, and an agent with a deontological view must be an agent feeling remorse over the killing of an innocent, even when he has overwhelming reasons to do this.
The reason why this is so is that having any other attitude toward killing means having an attitude that is guaranteed not to produce the best results in the long terms. When I write "best results" I am not affirming a consequentialist views of rightness, but I am referring to agent's own take of the goodness of the outcomes, that is, an assessment of the goodness of the outcomes of having a certain attitude in the long run, as viewed from the point of view of the agent, which is something any sensible non-consequentialist theory must allow. (Alternatively, one can rely upon a contractualist framework and sya that the attitude in question is not the sort of attitude that could reasonably be accepted by agents looking forward to find universal attitudes to guide human behavior.)
Both when the agent is , and when he is not, aware of this relation between remorse and his long term attitudes towards morality, this justifies feeling remorse for having to kill an innocent even when this is morally justified. Remorse, that is to say, is exactly what he ought to feel if one is the sort of agent that has the right, that is to say the virtuous, attitude towards killing.