Noncognitivism, 'ought', and uncertainty
Suppose there are no objective moral facts. It's tempting to think that this calls for a special semantics for moral language. Perhaps moral statements somehow express moral attitudes rather than describe the world. The trouble is that moral statements seem to behave like ordinary descriptive statements. Not only can we freely conjoin moral and descriptive statements. We can even use the same words – say, 'you ought to leave' – to express a moral attitude but also to report the implications of some contextually salient norms. It would be nice if we could use a standard descriptivist semantics for 'ought' statements even if we don't believe in objective normative facts.
This is what Bob Beddor suggests in Beddor (2023). He suggests that we should assume a noncognitivist psychology, but a descriptivist semantics.
The proposal is attractively simple. Start with the well-known semantics of 'ought' associated with Kratzer (1981):
There are two parameters here: the accessibility relation and the betterness ordering. According to Kratzer, these are generally supplied by context. Let's make this explicit:
If there are no moral facts, the contextual betterness order for moral 'ought' claims can't be an objective moral order. Let's assume the relevant order is given by the speaker's preferences. This yields a subjectivist semantics: 'Ought φ' reports that the speaker desires φ. We can make this more palatable by moving from the speaker's preferences to, say, the idealized joint preferences of the conversational participants (as in Lewis (1989), for example).
Next, add the following semantics for belief reports:
The interesting part is that the ordering is not shifted to S's belief worlds. If we replace 'c-best' by 'w-best', beliefs about what ought to be the case would get analysed as beliefs about one's own desires. We wouldn't have a noncognitivist psychology. (B) amounts to a kind of hybrid/relational noncognitivism: moral attitudes consist in a relation between descriptive beliefs and (actual) desires.
Beddor has more to say about all this, much of it interesting. I want to raise one problem and air a suspicion.
The problem I want to raise is that this proposal can't make sense of moral uncertainty. Suppose you know what you believe (which worlds are doxastically accessible), and you know your preferences. Then you know whether φ is true at the most preferred among your doxastically accessible worlds. According to (B), you can't be unsure about what ought to be the case. Moral uncertainty is only possible for agents who are unsure about their present beliefs and desires. This seems wrong.
Would it help to replace the ordering from the agent's preferences to something less accessible? What if we read 'c-best' in (B) as 'best by the shared preferences of the conversational participants in c'. Does this help?
No. It makes things worse. (B) now says that you believe that φ ought to be the case iff you believe that things are such that it would be best if φ by our actual shared preferences. You don't need to have any idea what these preferences are.
The failed revision illustrates that knowledge of preferences is beside the point. As long as you know what you believe, (B) implies that you can't have moral uncertainty.
So this proposal doesn't work.
Here's my hunch. Our semantics should be realist and objectivist. We should define truth relative to "contexts" that include a moral ordering. The semantics of moral 'ought' should pick up on this ordering. The semantics of 'believes that ought' should pick up on people's beliefs about the ordering. That is, we should give up not only expressivism about language, but also noncognitivism about moral attitudes. People can have genuine beliefs about moral issues. Their credences are defined over a product of descriptive and moral propositions. In a second stage, we now take this whole model of semantics and psychology and explain how it can be interpreted in a world without moral facts: how it can be projected onto a lower-dimensional space with only descriptive facts. Has this been tried?