These thoughts were engendered by some readings in Robert Nozick’s book Philosophical Explanations, in particular the discussion near the end of Ch. 5, Part III on “Deontology and Teleology”, and the preceding sections on the structure of moral rules. I do not think it is necessary to read or have read Nozick’s book in order to follow my meditations, though.
If you perform certain types of bad acts, even for a “good purpose”, you lessen the amount of good that you, and perhaps others, may do in the future. If you torture a known (even admitted) terrorist in order to thwart his plans and save innocent lives, you become the sort of person who will more easily torture in the future, perhaps sometimes on mere suspicion, and hence, eventually, an innocent. Also, you fill your victim’s relatives and friends with resentment, anger, hate, lessening the good they will do in the future, and making it more likely that they will do wrong. This doesn’t mean, necessarily, that one may never do a wrong act to accomplish a good end, but it is a factor to be weighed. I think this connects to some extent the deontological (rules-based) and teleological (ends-based) views of morals, and helping to avoid some of the worst “ends justify the means” abuses of vulgar forms of the latter.
Nozick’s ideas of the foundation of ethics have, I think, some serious flaws. The deepest flaw, to me, is his assumption (which he never really tries to justify?) of some Platonic realm of value, right, and wrong that transcends and has no necessary relation to (at least is not in any way derived from) human ends. I reject this view, and hope to set down some alternative speculations in some detail, in the future.
On a more technical level, his discussion of the structure of moral rules gives food for thought, but his elaborate formulation (which even he does not try to complete) is far too complex for actual application. Surely an analysis of morality must consider the “computability” of the resulting formulas – the possibility that the answers could actually be reached by real people in “real time” – otherwise it is asking us to be better than we possibly can be.
Perhaps deontological rules are best seen as heuristics (“rules of thumb”). Heuristics are designed for quick computations that give good (not necessarily optimal) results in many (hopefully “most”) situations. A set of heuristic rules does not necessarily need to be internally (logically) consistent. Judgment applies in deciding which heuristic to apply, or even whether to apply the heuristics at all, rather than opt for some more “precise” formula such as a careful, weighted analysis of the long(er)-term moral benefits and costs.