Tito Magri
Lecture Notes 2013
The Semantics of Expressivism
The most developed program, in the area of the semantic conditions for moral inference and reasoning, is that of the different forms of Expressivism. We have discussed Expressivism as a kind (possibly the most articulated and refined one) of moral anti-realism. I mentioned, the problems that arise for Expressivism in relation to the semantic and inferential relations of expressively understood moral claims. Now the moment has come for discussing these matters, since contemporary Expressivism has done a lot of work to ground the possibility of moral reasoning on an appropriate conception of content.[1]
The general problem raised by moral reasoning is to identify, and to connect in the appropriate way, the moral aspect of content that is preserved in inference and the formal conditions that ensure its preservation. The notions of truth and validity provide an answer in the theoretical case. The task is to find out whether counterpart notions can be found for moral reasoning. Of course, all varieties of first-order moral realism are in position to predicate moral claims of truth or falsity – after a minimalist fashion. But this is far from solving the problem. On this view, truth and falsity of moral claims are first-order predicates that rely on the independent availability of a conception of what is to rationally engage in first-order moral practice and discourse. The philosophical program of contemporary Expressivism is to give an explanation of the truth- and justification-aptness of moral claims that is itself thoroughly moral, that proceeds by identifying in various ways, at all levels, conceptual, epistemological, metaphysical conditions on moral thought and discourse with conditions on sound, good, rational morale practice - that are themselves not understood as the recognition of certain truths or facts. This takes the general form of an account of what moral claims mean, of the meaning of the predicates that allow forming judgments of a moral sort, and of their other semantic features in terms of what is done, of the kind of action that is performed by means of such claims and in terms of the mental states the subjects who engage in such actions. The proposed explanation is pragmatic all the way down, the notion of claim being itself pragmatic: it is a sort of expressive speech-act. What is expressed in a claim is a stance or commitment, and thus the expressing itself is a kind of activity which involves a mental state of the agent. The task is thus incumbent on Expressivism of explaining, starting from the resources offered by a non-conceptual understanding of the mental stances underlying moral claims how these manage to behave like ordinary descriptive claims and also their bearing on inference. The task (in Blackburn’s apt phrasing) is that of «earning the right» to truth and to reasoning in connection with moral discourse. This is a task that strictly arises only for Expressivism. It is doubtful how well Expressivism can manage to perform it; but it is certainly a highly generative philosophical program.
EmbeddingThe central problem for Expressivism, in this regard, is the Problem of Embedding. Let suppose, with Expressivism, that the content of a moral claim is a certain function of what is done in making that claim. Let suppose, also, that the relevant speech-acts do not already have content of a sort that grounds their correctness and inferential relations. Rather, speech-acts have explanatory weights simply as doings that are ultimate either in their impulsive experiential character or in their pragmatic-functional role. And their contents are not determined as propositions but rather are a function of mental states.[2] But a problem is raised by the (strictly and strongly) pragmatic and mentalist character of this account. If moral content is completely determined by the pragmatic significance of certain doings (claiming something, endorsing something), it is not clear how we are to account for embedded moral contentsthat are pragmatically neutral. If there is nothing more to moral content than what is provided by this derivation from the pragmatic force of an episode, more precisely, if moral content is, at its root and in its essence, not content but linguistic practice and mental activity, then there is a difficulty about non-assertive (generally, non-committive) contexts necessarily arises. But with this difficulty goes that of establishing something like a «logic of attitudes».[3] To see why a logic of moral claims is difficult to obtain for Expressivism, we may concentrate on a presumptive form of moral argument (that is, whose premises and conclusion ostensibly have moral import), one, for instance, that has the appearance (and the rational allure) of modus ponens:
(1)One ought to honor his father
If one ought to honor his father, then one ought to honor his mother
------
One ought to honor his mother
This seems to be a very good argument, one we should undoubtedly accept. That is, it seems to be valid, so that, if the premises are true, so the conclusion is. But, now, we should raise the question: Is this argument truly a case modus ponens? This is to ask whether it recommends itself, as a step to take in reasoning, because and just because it exemplifies the scheme
(2)(p & (p →q)) →q
Modus ponens consists in a syntactic structure that, in turn, makes certain minimal semantic demands on its components. Thus, for an argument to be recommendable as a valid instance of modus ponens, it is required that, in (2), p occurs in the two premises with the same meaning. But suppose that the meaning of p is given through an expressivist analysis, according to which what is meaningful, and imparts meaning, is the making of the claim that p as this is expressive of the actual mental states of the concerned subject. Now, the occurrence of p in the second premise is embedded in the antecedent of a conditional. Thus, such occurrence cannot count as the making a claim (what is claimed is the conditional, and this does not entail that the antecedent is claimed). In effect, it does not at all count as an act, thus, a fortiori, as an act of expressing an occurrent mental state. On the expressivist account, then, p does not have the same meaning in its second as in its first occurrence (which, by contrast, amounts to a claim) in (2). Possibly, in its second occurrence, p has no meaning at all, since the pragmatic, expressive requirements for meaningfulness seem not to be met. The consequence of this seems to be that (1), under an expressivist analysis of its components, is not, against all appearances, valid as an instance of modus ponens. It fails for equivocation in the premises. This is the Problem of Embedding, or the Frege-Geach Problem.[4] The problem is easily generalized, for Expressivism, to other inferential schemes, given only that they include embedded and non-embedded occurrences of a proposition. So
(3)p
------
p v q
is not valid, under an expressivist account of the meaning of p, since to claim a disjunction is not to claim any of the disjuncts. By contrast,
(4) pq
------
p
is valid, since to claim or assert a conjunction is to claim or assert each of the conjoints. But, in effect, given the connection between conditionals and implication, we can safely limit our discussion to the case of modus ponens. An even more compelling theme would be that of negation, since Expressivism meets serious difficulties in explaining how moral sentences which are one the negation of the other – “Murder is wrong” and Murder is not wrong” – are inconsistent the one with the other. And inconsistency is the key to logical consequence: a sentence logically follows from others is its negation is inconsistent with the negation of their conjunction. But I will not address here this issue.[5]In general, unless something is done, Expressivism cannot account for the validity of various inferential schemes and for the general relation of logical consequencewith reference to moral claims. But, on the other hand, it seems that moral or prudential reasoning is a fact and that it cannot be divorced from standard, valid inferential schemes, nor fail to instance the relation of logical consequence among its constituents. Unless we accept that, by embracing Expressivism, we are deprived of any ground to construct, and assess, arguments from, and for, moral authority, the Problem of Embedding has to find a solution.
Two ApproachesContemporary Expressivism has dealt with this task along two, strictly related but quite distinct, lines of argument. One consists in the program of framing a syntactic structure for attitudes (Logic of Attitudes), the other focuses of the idea of a consistent set of attitudes (Normative Logic). There are, obviously, common threads - besides the general expressivist tenet. The prominent one is that, given the essential connection between validity (necessary truth-preservation), logical consequence (the inconsistency of the set of premises and the negation conclusion), and truth (which is involved in the definition of validity and logical consequence), the task of providing for logical relations is inextricably linked with that of explaining content. But the two programs diverge in interesting ways.[6]
The Problem of Embedding can be discussed from a fundamentally syntactical or from a fundamentally semantical perspective. It can be seen as the problem of how Expressivism can account for the patterns of relation among claims and contents that make possible appropriate (if logicality is attained, valid) transitions of thought. Or it can be the problem of how Expressivism can account for the normative property of claims whose preservation makes for the inferential import of thought-transitions (in the logical case, validity). The problem can thus be viewed as one of normative constraints on content that spring from inferential patterns, or as one of normative inferential patterns that are made possible by properties of contents. On the first view, the task is that of providing a non-truth-functional interpretation of logical connectives, so as to extend correspondingly the scope of inferentially valid schemes, and thus make sense, without assuming truth-aptness from the start, of context-invariant properties of content. On the second view, the task is to show that pragmatically determined attitudes can stand in certain primitive, non truth-functional, relations of content, that make sense in a uniform way of the conditions of consistency of sets of norms, and thus give an interpretation to valid inferential transitions among claims without cognitive import. Another, very recent approach is that of addressing both the logical and the semantic issues from the standpoint of the requirements of the semantic of a moral fragment of natural language; and of assessing from this standpoint whether there is a mandatory interpretation of moral expression, inference, and attitudes. But I will not take this up for discussion. [7]
Layered Stances The program of a Logic of Attitudes is syntactically oriented (that is, gives priority to the account of logical connectives and patterns of implication, rather than to properties and relations of content) and, in a certain respect, more markedly pragmatic than the alternative, normativity-oriented one. Its original move is that of turning apparently non-asserted, embedded, occurrences of purportedly moral contents into asserted ones, into commitments to those contents. This is obviously a pragmatic move.[8] The idea is that the syntactic structure underlying inferences from or to moral conclusions, reflects combinations of lower-level attitudes (coming to expression in the inferential constituents) and of higher-level attitudes, taken towards those combinations. More precisely, inferences of such a kind get their normative fabric (and thus conditions for its validity, and for whatever force they can have) from the fact that moral stances can be taken toward combinations of moral stances, astance of acceptance or refusal toward having them together. The latter stances, in turn, can consist (like the second premise of (2), above) of narrower-scope combinations of stances and of related higher-level attitudes, all the way down to individual, lowest-level, constituents. This nested, roughly recursive, structure of first-order commitments comes to expression in the logical syntax of inference from and to moral claims. To endorse or to reject having certain attitudes together is to treat them in a way that expressively comes out as consistency and inconsistency. And if the set of attitudes that occur as constituents in a certain pattern of commitments (some of them involving the others) is consistent, then such a pattern licenses logically valid transitions among them. This view may be supported by facts of our moral experience. It is a fact that moral stances and commitments tend to occur as more or less regular complexes, making for recognizable, generative patterns of behavior and characters - what might be called sensibilities.[9] It is also a fact that, given the general functions of moral authority, attitudes taken toward sensibilities are significant and important. Suppose such an attitude is taken against the sensibility that consists in combining attitudes so as to disrupt a regularity of the kind exemplified in (2). The ground for this could be the «clash of attitudes» that such sensibility comports, its being a «fractured sensibility» that cannot fulfil the general purposes of practice nor a fortiori be invested of authority. This would come to expression in normatively assessing such sensibility as if it involved endorsing the premises and rejecting the conclusion of moralmodus ponens. The blame of being inconsistent in one’s moral or prudential commitments - a moral condemnation of a particularly serious kind - would take us to reject as invalid a step of moral reasoning.[10] This may be held to provide non-cognitive, non truth-functional, counterparts to syntactical structures permitting or forbidding propositional combinations and inferential transitions, that is, non truth-functional counterparts to consistency (approval of a combination of attitudes) and to validity (the inconsistency of approval of the premises and rejection of the conclusion).[11]
A Logic of Attitudes The logic corresponding to this view aims to specify of the rational relations of moral claims by proceeding from the outside to the inside.[12] Higher level, first-order commitments form the syntax of the complexes of lower level attitudes that they take as their objects and assign meaning to the connectives holding them together. With this structure in place, the expression of attitudes is constrained by its taking place in the context of an activity of framing and assessing moral arguments. Within this structured practice, and the roles and the pressures it embodies, individual attitudes come to expression “as if they were judgments” and, consequently, as if they advanced claims to justification and truth. This, of course, does not leave unaffected their contents, rather confers to them the objectivity that amounts to (quasi-realistically) “projecting” attitudes, so as to make sense of their appearing to respond to features of the world, and of the standard, quasi-truth-conditional interpretation, of the relevant predicates.[13] A Logic of Attitudes is a part, rather than a consequence, of the general quasi-realist program. If this pragmatic account of logical consequence were really successful, the Problem of Embedding could be easily sidestepped. Ascription of content to moral claims, on this expressivist proposal, takes place only modulo the higher-level, first-order pragmatic rendering of inferential roles. In other words, the general kind of commitments, involving (inter alia) p, that come to expression in (2), or in other potentially inferentially relevant contexts involving p, contribute allover to the meaning of p. Therefore, there is no principled distinction between pragmatically neutral and pragmatically committed occurrences. Rather, we have always to do with commitments having different scopes and different, but interrelated, structures. Determination of content, thus, can be uniform through asserted and non asserted contexts, just because all contexts (governed by logical connectives) are, at some level, context of moral commitment, and the determination of content for attitudes is governed from outside in, that is, form higher to lower level. This seems to make available an expressivist explanation of meaning that is uniform across syntactically and pragmatically different contexts, and that underpins inferential relations among contents.[14]
ProblemsHowever, the program of a Logic of Attitudes is beset with serious problems. The program can be developed in different ways. The main divide is that between replacing logical constants (by giving a directly moral meaning to sentential complexes linked by connectives, without relying on their truth-functional interpretation) and retaining them (and giving an appropriate interpretation of attitudes embedded in sentential compounds constructed by means of ordinary connectives).[15] Even bracketing all technical issues, the first alternative (the one I have been following in my presentation), that puts higher-level attitudes directly into the content of conditionals and syntactic compounds, seems to be confronted with a general difficulty. It has been repeatedly observed that in this way moral arguments could not be established as logically valid. A clash of attitudes, fracturing our sensibility, can be a moral shortcoming, can defeat the governance of conduct and even the possibility of practice. But it does not count as a logical fault. There can be merits in reasoning along the lines of moralmodus ponens, and demerits in not accepting its conclusion while embracing its premises. But they are moral, not logical ones. Thus, moralmodus ponens cannot be established as a logically valid inference.[16]
On the alternative proposal, moral conditionals, or other compounds, result from combining moral commitments (and also other attitudes, like beliefs) and embedding them in complexes governed by ordinary logical connectives. Higher-level attitudes are not determinants of the meaning of connectives, but are part of the explanation, or of the reason, for the shape of one’s overall moral position, whose logical form is what should be made explicit.[17] This task consists in showing that the truth-functional meaning of connectives can be preserved, even under an expressivist interpretation of the elements that they combine.[18] This is attempted by applying to all such compounds and inferential complexes (in normal form) the method of tableaux, so as to make explicit their structure and follow their implications, and by proposing an expressivist semantics for attitudes, so as to make available a notion of consistency that can be applied in testing inferences for validity.[19] (Thus, in the case of moral conditionals, one is “tied to a tree”, to a disjunctive commitment, that is, to endorse one branch, if he cannot endorse the other. This instructs about how to check consistency.) The main idea underlying the proposed semantics, and rendering of the notion of consistency in attitudes, is following sets of attitudes through their realization in possible situations, in order to test whether they can be jointly satisfied.[20] It is interesting to observe that, when it comes to characterizing the semantic properties of commitments, their relation to the world, specified in terms of the situations that do, or do not, count as realizing them, comes to the fore.[21]