Suppose that a world in which we have an utterly non-consequentialist moral status is a better world than one in which we don't have such a status. Does this give any reason to believe that we have such moral status? Suppose that a world without moral luck is worse than a world with moral luck. Does this give any reason to believe that there is moral luck?

The problem is that positive answers to these questionsFootnote ¹ seem to commit us to instances of the inference ‘Wouldn't it be nice if p, therefore, p’. Perhaps it would be nice if we had utterly non-consequentialist dignity. How is this any reason to believe that we have such status? Perhaps it would be nice if there were moral luck. How is this any reason to believe that there is moral luck? Thus stated, such inferences look ridiculous, paradigmatic cases of wishful thinking.Footnote ² And yet they do not sound so obviously ridiculous, at least not as ridiculous as non-moral instances of the same argument schemaFootnote ³ (it would be nice if there were world peace, therefore, there is world peace). Can something be said in defence of such arguments, at least in morality?

One way of answering in the positive would be to endorse a not-completely realist metaethics. If, for instance, some mutually beneficial contractarian scheme determines whether or not we have the special non-consequentialist kind of dignity, then the argument may work: a world in which we have such a status is a better world, so it is in our interest to see to it that we have such a status, so a mutually beneficial contract would see to it that we have such a status, so – given that such a contract determines whether or not we have such a status – we do have such a status. Similarly, the argument can be made to work if, for instance, whether or not there is moral luck is determined by the will of an omnibenevolent creature: if a world with moral luck is better than a world without moral luck, then an omnibenevolent god would see to it that in the actual world there is moral luck. And since it is – we are now assuming – up to her whether or not there is moral luck, we can safely conclude that there is moral luck.

But such replies are based on controversial – and wrong, I would say – metaethical views, and it may anyway be desirable to find a more general way of vindicating ‘Wouldn't it be nice if p, therefore, p’ inferences in morality.

Some arguments of roughly this form are, I think, valid. Some, of course, are not. I suggest that the question whether a given argument of this form is valid is best reduced to another question, one about the logical nature of the operators involved, and in particular whether they satisfy the analogue of the (suggested) modal-logic axiom ◊◊p→◊p.

Some operators clearly fail to satisfy such an axiom. Consider, for instance, ‘I desire that I desire that I only eat low-fat food’, which sadly does not entail ‘I desire that I only eat low-fat food’ (at least if this is understood as ‘I desire to eat only low-fat food’). With some operators, such iterations are not even intelligible, perhaps not even well-formed. Consider ‘It is healthy that it is healthy that I eat only low-fat food’. But some operators – and indeed, some normative operators – do seem to satisfy this rule. Perhaps good is an example: ‘It is good that it is good that I watch my weight’ does seem to entail ‘It is good that I watch my weight’, at least with the meaning of ‘good’ held fixed throughout. (So it is not a counterexample that it would be fun-good if it weren't health-good that I watch my weight.)

For operators of this sort – operators, like perhaps ‘it is good that’, that satisfy the GGp → Gp rule – something like the argument schema I started with can be made to work. If, for instance, the debate over our dignified status can be understood as a debate over something of the form it is good that q (for some q), then the inferential move from it would be good if we had a non-consequentialist dignified status to we have such a status could be vindicated as an instance of GGq → Gq. Similarly for other moral debates (like the one about moral luck) and other operators, so long as they satisfy the analogue of the GGp → Gp rule.

So some arguments roughly of the form ‘Wouldn't it be nice if p, therefore, p’ may be valid for a moral p. Such arguments can be neither accepted nor rejected en bloc.Footnote ⁴ Rather, what the proponent of such an argument owes us to vindicate her (suspicious) argument – and what the opponent of such an argument must give reasons to reject – is a defence of the following claims:

1. (1) p ≡ Gq (for some operator G, and some proposition q)

2. (2) The G operator satisfies the GGp → Gp rule.

Even then, what would be vindicated is not exactly the argument ‘Wouldn't it be nice if p, therefore, p’, but rather, assuming that G stands for ‘Wouldn't it be good that’, the argument ‘Wouldn't it be good that p, therefore p’. But this, it seems to me, is close enough to what proponents of such arguments are after.Footnote ⁵

Let me not overstate the progress that has (perhaps) been made here. I still haven't offered any way of determining – for a given operator – whether it satisfies the GGp → Gp rule. But some progress has been made, because we started with what seemed on reflection not obviously bad arguments which are nevertheless instances of obviously fallacious argument schemas, and an intuition that there's something peculiar to morality that licenses such arguments within it. And my suggestion here – to understand and evaluate such arguments as instances of a GGp → Gp inference, for some operator G – explains and vindicates our starting point: it is not a generally valid inferential rule, but for some operators – and perhaps especially moral ones – it may very well be.Footnote ⁶