Suppose that we have a complete consequentialist ethical theory, and from that theory we have derived a set of rules. We have chosen these rules because if we follow them the best outcomes become most likely (following the rules is justified by our consequentialist ethical position that we should act so as to achieve the best outcome). One such rule is probably as follows: don’t break the agreements you enter into with other people. Such a rule apparently defeats one of the classically problematic cases for consequentialism, namely whether it is right for a doctor to kill one healthy man against his will in order to save five others. Consequentialism by itself seems to imply that we should kill the healthy man, because it is better to have five people alive and one dead than the other way around. Clearly this is contrary to our ordinary ethical intuitions. But if we are following our set of rules in order to achieve the best outcomes we will not endorse killing the healthy man, because that would violate our agreement with him (implicit upon admission to the hospital) to try and make him healthier. And, if following the rule seems contrary to consequentialism, we can point out that if we didn’t generally respect such agreements people would be afraid to go to the hospital, and so the rule really does ensure the best outcomes overall.
But nothing in the considerations above motivate total obedience to the rules. We are only following the rules so as to achieve the best outcomes overall. This raises the possibility that we might on occasion be justified in breaking the rules, if we can achieve a better outcome by breaking them than by following them. Perhaps we can kill the one healthy man secretly, in such a way that it doesn’t have repercussions for the reputation of the hospital. In such a situation it would seem that we could achieve the best outcome by breaking the rule in this specific case, while following it elsewhere, and thus achieve the best possible outcome. And of course this is just as much a conflict with our intuitions as before.
However if we carry that line of reasoning out then it would appear that in every situation we should consider weather we can improve the outcome by deviating from the rules, and that when we can, we should. But this makes the rules superfluous, because we are back to considering what the best outcome is in each particular situation by itself. Clearly we have made a mistake in our reasoning somewhere.
To understand how we have gone wrong we need to take a detail to a different kind of rule, statistical prediction rules (SPRs). An SPR is a mathematical formula that takes a number of quantifiable facts as inputs and yields a predication as to how likely an event is to occur. For example, an SPR might be devised that takes the amount of rainfall, average temperature, and the quality of last years harvest and predicts whether the harvest will be good or bad. Despite using only a limited amount of information these SPRs are on average better than even experts when it comes to predicting the outcome of events. Now consider trying to make a prediction given an SPR. The SPR says that the event probably won’t happen, but your best reasoning says that the event will happen. Can you improve your prediction by overriding the SPR? On average you can’t, that is what saying that the SPR is better at predicting the outcome of events means, usually when you disagree you are wrong.
Our decision to deviate from ethical rules is like a decision to deviate from an SPR. Ethical rules are created so as to yield the best outcome in the most possible cases, better than reasoning about each individual case on its own merits would (at least if they are justified). Thus deviating from the ethical rule will, on average, result in a less than optimal outcome, even if it seems to you that deviating from the rule will produce a better outcome. We aren’t perfect when it comes to determining the outcome of our choices, this is why we make ethical rules.
This doesn’t mean we should never break the rules, but it does change when we should break the rules. Again, going back to SPRs, we can beat the predictions of the SPR if we are able to take advantage of information that has a strong influence on the outcome of events that we know the SPR does not take into account. For example, if we know that someone has broken their leg we may be able to beat the SPR at predicting some of their actions, assuming it doesn’t take that injury into account. So, in terms of ethical rules, we should only break them when it seems like factors that the rule was not designed to take into account have a strong influence on the outcome. For example, our traditional ethical rules regarding “theft” were obviously not designed with the possibility that something could be “stolen” without depriving someone else of property, and so in the case of file sharing we may be justified in breaking this rule.
But of course these considerations all assume that we can actually create ethical rules that we do better on average by following than by relying on our own judgment. If we can’t create rules of this caliber then obviously all bets are off, and it would instead to better to do without rules completely. Now I have no direct evidence that such rules can be fashioned. However, the fact that cultures do in fact contain rules like this and lean on their members to obey them implies that following rules can guarantee better outcomes than relying on our judgment, otherwise more successful cultures would have developed that instill in their members a sense of what the best outcomes are, and lean on them to achieve them directly. If ethics is to the benefit of society, which I think that it is, then this is indirect evidence that following rules can lead to better outcomes than judging situations on a case by case basis, in general, and thus that we can apply all I have said above to such rules.