Keeping a rule against the case

Why should you keep a rule in the one case where breaking it would plainly do more good?

Here is a situation everyone recognises. There is a rule — keep your promises, do not lie, do not open another person's mail — and here is a case where breaking it would obviously produce a better outcome. Not a marginal case: the harm of keeping the rule is plain, and so is the benefit of breaking it.

The question is not whether you feel a pull to keep the rule anyway. Most people do. It is whether that pull is more than squeamishness. If the rule exists because following it makes things go well, and here it makes things go badly, what could justify following it?

Begin by noticing that the case has been described from a strange vantage point: yours, now, with the calculation already made and known correct. How often is that vantage point real?

You are estimating consequences that have not happened, on incomplete information, at a moment when you have an interest in the answer. Almost every consideration favouring a breach arrives in a form indistinguishable from motivated reasoning — and the person about to make a self-serving exception feels, from the inside, exactly like one who has correctly identified a genuine one. So a rule earns its keep partly by taking the decision out of the hands of the least reliable judge available, which is you at the moment of temptation. That is R. M. Hare's two-level picture: at the intuitive level we run on firm dispositions, because that is where agents under time and self-interest perform best, while at the critical level, in calm reflection, we ask which dispositions to cultivate. The rules are justified by consequences; they are not applied by calculating consequences.

A second consideration is worth keeping separate. Some rules produce their benefit only by being reliable. A promise is useful because it lets someone plan; a court's rulings guide behaviour only if the next court rules likewise. The value there lies not in each act alone but in the pattern, and a defection that is individually harmless still spends a little of the pattern's credibility. John Rawls pressed this further: where an act is defined by a practice, as promising is, appealing to the good consequences of this particular breach is not a permitted move within it.

The law shows the same structure without the moralising. A rule bars vehicles from the park, and here is a case — an ambulance, a war memorial jeep — where its purpose is not served. A formalist judge follows the text anyway. Frederick Schauer's account is that rules are deliberately entrenched generalisations, built to be a little wrong in particular cases, because their function is to settle matters in advance; a rule that yields whenever someone shows it producing a poor result here is not a rule but an invitation to argue.

So we have three reasons, and they are not the same: agents judge badly in the moment, some goods exist only in the pattern, and settled expectations have value of their own.

Now the objection. J. J. C. Smart named it: if the rule is justified by its consequences, and you can see that this act's consequences are worse, then keeping it is rule worship — treating the instrument as the goal. Notice that the first defence is contingent. It says you probably cannot tell that this is a genuine exception. But sometimes you plainly can, and the moment you can the defence has spent itself, which turns the two-level view back into act consequentialism wearing a habit.

Brad Hooker's rule consequentialism answers differently: we assess a whole code, chosen for the consequences of its general acceptance, and the right act is what the best code permits even when this act is not the best available. That is coherent, and it collects Smart's charge in a sharper form — the theory now recommends a suboptimal act while resting on the value of optimising, and nobody has closed that gap. Rawls's practice conception dodges it, but only for acts a practice constitutes; it does not cover a rule like "do not lie", which you can break without abolishing anything.

The three defences hold to different degrees. The unreliable-judge one is empirical and weakens as the case gets clearer. The pattern one applies only to rules whose good really is produced by regularity. The reliance one is strongest where others are planning on you, and near zero in a secret one-off.

So the honest answer is conditional. Rule-following is best supported when your judgement is untrustworthy, when others are relying, and when a wrongly-made exception costs more than a wrongly-missed one — and weakest in the clean, secret, non-repeating case that philosophers construct for exactly that reason.

And the position is not "never break a rule". It is that the threshold for departure should sit well above "this looks better to me", because at that threshold the rule has no force at all. Where above it the threshold belongs nobody has derived; it is judged, which is uncomfortable for a view whose appeal was that it removes judgement.

• How threshold deontology tries to fix a level above which a rule yields.
• Why legal systems keep equitable powers alongside formal rules, and what that concedes about formalism.

Every term the collection defines is gathered in the glossary.