Having defeated expressivism (see here for his arguments) in a way that suggests the necessity of a truth-conditional semantics for normative statements, Wedgwood outlines what a factualist semantics must accomplish, and in so doing, reveals the failure of two factualist semantic theories: the causal theory put forward by the Cornell realists, and the conceptual analysis theory put forth by the Australian realists.
Requirements for Factualist Semantics
Propositions are made true by the actuality of the properties being in the described relation in which they are stated, such that normative terms refer to normative properties and the semantic value of those normative terms. Propositions must also be understood as being built up of concepts that represent their respective semantic values, in such a way that concepts are modes of “presentation” of semantic values. When we make a moral judgment both propositions are at play – a thought that is built up by the concepts, and the conditions for which that thought would be true, built up from the semantic values of those concepts. Our semantic theory must offer an account for the possibility of both sorts of propositions, which will include an explanation of what an understanding of the concept will consist in, and how the concept will refer to the semantic value of that concept. As we are dealing with normative concepts, our explanation of the essential nature of a concept that amounts to the understanding of that concept must address the very normativity of that concept. It is this last component of a semantics of normative terms that Wedgwood posits cannot be met by causal accounts and conceptual analysis accounts.
The causal theory of normative semantics holds that normative statements posit normative concepts that are caused in the right sort of circumstances by normative properties. In other words, for a judgment to be a moral judgment is for it to involve an ought-concept that is reliably caused by an ought-relation. Wedgwood notes that those philosophers who are skeptical of this position tend to do so on the grounds that it is false that normative properties and relations are causally efficacious (see here for Harman’s famous objection, and here for Audi’s rejection of the causality of moral properties), as this view requires, but Wedgwood is skeptical of this view not because normative properties are not causally efficacious, but because a causal semantics allows for the possibility of normative relations that are not motivating in a way that is inconsistent with internalism.
To make this point, Wedgwood indulges in a fun thought experiment, wherein an intelligent android is designed to offer highly reliable advice about what to do in moral affairs. By accident, the android is marooned on an alien planet, and the aliens there come to understand all the terms of his language, save for “ought”, which they come to consider as referring to whatever relation is causally responsible for his using that term. Over time they come to recognize actions that are consistently of the sort for which the android posits “ought-ness”, such that they can guess with a great degree of confidence whether the action they are considering has the property that the android refers to as “ought”. It seems then, that the aliens will have a concept that is caused by the normative property of “ought-ness”. Nevertheless, it seems quite possible that though their concept is causally regulated, through the progress of time, by an ought-relation, that their judgments involving this concept might not be motivating. If this is possible, then the causal theory is insufficient in that it does not secure the intrinsically motivating character of moral concepts. This is a variation of Horgan and Timmons moral twin earth argument (see here for more on Cornell realism and this objection), and while I have had reservations about H&Ts argument, Wedgwood’s adaptation of it seems convincing.
The conceptual analysis approach, whether in its reductive treatment by Frank Jackson and Philip Pettit, or its non-reductive treatment by Michael Smith, also fails to account for internalism, a failure which is a greater strike against it than it for the causal theory, because while the causal theorists deny the truth of internalism, proponents of conceptual analysis developed it to explain internalism, such that it fails on its own terms.
In general, the conceptual analysis approach takes it that a universally quantified biconditional can be offered where on one side there is a normative term, and on the other is the disjunctive set of natural terms that fix the extension of that normative term, such that conceptual analysis explains the semantic value of a term, because we have implicit knowledge of this biconditional through common moral platitudes. Wedgwood notes that this thesis is dubitable due to its resting on the notion that such identities are necessarily true, but pursues a different objection, by arguing that even if conceptual analysis can elaborate the semantic value of a normative term, it does not fully explain the concept the term expresses, because natural terms cannot capture the very normativity of the normative concept.
To make this point, Wedgwood appeals to Moore’s open question argument, by noting that if the conceptual analysis approach is correct, then if one knows that the facts that make true one of the disjunctives on the natural term side of a conceptual analysis obtain, then one has implicit knowledge that the normative term obtains, such that one must be motivated to act in accord with the internally motivating content of the concept that term expresses. But it seems that we can know this and not have the intention to act, such that such a reductive analysis of normative terms does not succeed because it would entail the possibility of making a normative judgment without having the intention to act in accord with that judgment, which is the negation of internalism.
Michael Smith’s non-reductive take on conceptual analysis does not fare better, because it only shifts attention from normative terms to rational agents. Smith’s view is that normative terms reduce to summary points about more specific platitudes of practical rationality, such that when we say that we ought do x, we are positing that if we were fully rational we would want that we do x, where “fully rational” entails having all relevant true beliefs, correct deliberation from those beliefs and other platitudes. The reason this analysis is non-reductive is because the notion of a fully rational agent is specified as one with the most unified profile, which is a normative notion.
Wedgwood’s objection to Smith’s theory is that even though it appears to account for internalism, it in fact fails to do so, because it only describes the phenomena of internalism – that is, seeing that one would be motivated to act if one had a reason to so act – it does not entail that internalism is a conceptual and necessary truth that follows from the function of normative judgments (see here for Wedgwood’s more in depth arguments for internalism). From all that has been said, an agent may see no reason to work towards the most unified profile, as the only reason to favor a more unified profile than a non-unified profile might be because the unified profile is “prettier”. If this is so, then internalism rests on the contingent factor of a desire to have this most unified profile, and as internalism is a necessarily true thesis, this account of internalism falls short.
The lesson to take from the failure of the causal account and the conceptual analysis account is that an appropriate account of moral semantics must address the distinctive action guiding role that normative concepts play in thought, which suggests that a semantic theory will not succeed if it does not include normative terms in the metalanguage. So the conclusion we are left with is that the causal account was onto something, in that it did allow normative terms into the metalanguage by specifying that normative relations causally regulate the use of normative concepts and terms. As such, normative relations can be featured in the ground level of our semantic theory; a normative term cannot be used to explain its own meaning, but we must be able to talk about normative relations in order to convey how terms can express those relations.