At the level of descriptivism, there is an obvious structural parallel between the landscape of metaethical and metaepistemological positions. Universalism, speaker-relativism, and subject- relativism—in metaethics—map fairly closely onto invariantism, contextualism, and subject- sensitive invariantism—in metaepistemology. Moreover, some of the challenges dogging each pair of parallel accounts are similar. This parallel is not exact, but even the inexact parallel and the thought that ethical and epistemic claims are both normative sorts of claims encourages exploring the possibility of nondescriptivist accounts of epistemic claims parallel to the nondescriptivist accounts of ethical claims explored in previous chapters. I suspect that this has not been done because metaepistemologists have assumed that the only possible nondescriptivist account is an analogue of the traditional and crude emotivist form of noncognitivism defended by Ayer and Stevenson, yet such a view of knowledge claims is patently implausible. But, in my view, emotivism is implausible both when applied to epistemological discourse and when applied to ethical discourse, for all of the reasons discussed
in chapter 2. Yet surely the failure of emotivism doesn’t undermine all attempts to develop a form of ethical or epistemic noncognitivism. In this section, I want to sketch and defend what I take to be the most plausible version of epistemic noncognitivism. In effect, it will be an extension of Gibbard’s quasi-realist version of expressivism.
As we’ve seen, the most worked out version of ethical noncognitivism is due to Gibbard (1990), who argues that ethical claims mean what they do in virtue of their expressing a state of norm-acceptance. His account of norm-acceptance can be viewed as a simple but powerful modification of the relativist view that the truth conditions of ethical claims are implicitly relative to a set of ethical norms. In one sense, all descriptivists about ethical claims can agree that the sentence,
(1) George’s starting the war was immoral, has truth conditions given by,
(2) George’s starting the war was forbidden by ethical norms N.
The universalist will say that there is only one N for all tokens of (1), while the subject-relativist will say that N varies depending on features of George and the speaker-relativist will say that N varies depending on features of who is saying (1). Gibbard’s view can be seen as a modification on this speaker-relativist view. He argues that ethical claims express(p) an interlocking composite
mental state called norm-acceptance. This state comprises a descriptive belief whose content is (2) and a state of accepting the relevant norms N.
In chapter 3, we saw the advantages and disadvantages of this position in metaethics. What I want to do now is use this as a model for a version of epistemic noncognitivism. We’ve seen that speaker-relativism is parallel to epistemic contextualism. Contextualism, recall, can be summed up as the claim that sentences of the form
should be understood as having the truth conditions given by, (4) S’s true belief that p meets epistemic standards e,
where the value of eis determined by features of the speaker’s context. Rather than explaining the meaning of instances of (3) in terms of expressing a proposition with relativized truth conditions given by (4), as the contextualist does, I suggest that an epistemic noncognitivist could treat instances of (3) as expressions of states of epistemic norm-acceptance. It is epistemic norm-acceptance in that the norms accepted are epistemic norms, but norm-acceptance is, again, a complex state of mind consisting of a belief with (4) as its truth conditions and the acceptance of the epistemic norms alluded to therein.
When it comes to (at least some) ethical claims, it is pretty clear what the relevant norm is; they tell us that certain actions are obligated, permitted, or forbidden. But, although it is often said that epistemology is a normative discipline, it is less clear what epistemic norms are. A plausible suggestion here is that the norms relevant to understanding the meaning of knowledge claims are norms which entitle beliefs. Specifically knowledge claims could be understood as expressing our acceptance of particular epistemic norms, which when applied to a particular person’s belief entitle or don’t entitle the belief. So we might refine (4) with:
(4') S is entitled by norms eto her true belief that p.
So the proposal is that instances of (3) express a complex state of mind which includes a factual belief with (4') as its truth conditions and the acceptance of the epistemic norms ealluded to therein.
Is this norm-expressivist version of epistemic noncognitivism at all plausible? The best way to start to address this question is to see how well the view fares against the three challenges raised in our discussion of descriptivist accounts of knowledge claims. Because epistemic norm- expressivism basically incorporates contextualism into its account of the mental state expressed
by knowledge claims, it can claim many of the advantages of contextualism. But, I think, it also avoids, at least to some extent, the correlated disadvantages.
The primary advantages of contextualism were that it could meet Austin’s challenge by underwriting and explaining the apparent cross-context variation among ordinary speakers in willingness to attribute knowledge, and that it could meet the challenge of the paradox generated both of the AFC-paradoxes. At both points, however, the contextualist was forced to posit unexplainable forms of semantic blindness and thereby open himself up doubly to the challenge of semantic blindness. (See page 178f. above for precise articulation of these challenges.)
Regarding Austin’s challenge, epistemic noncognitivism can also explain the apparent cross- context variation among ordinary speakers in willingness to attribute knowledge. Recall DeRose’s (1992) bank cases, where, in a context where not much hangs on it, someone says that she knows that the bank is open on Saturday, but, in another context where a lot hangs on it but she has the exact same evidence, she denies this knowledge. A proponent of the sort of epistemic noncognitivism that I have been sketching would explain this variation much like the contextualist explains it—viz. in terms of a variation of the ein the content of the belief, which is part of the state of norm-acceptance. However, the epistemic noncognitivist will attribute this phenomenon not to cross-context variations in the operative epistemic standards but to cross- context differences in which epistemic norms are accepted by those attributing and denying knowledge. This can, in turn, be explained in terms of contextual variations in our reasons for evaluating whether or not a particular believer is entitled to his beliefs.
Contextualism achieves its answer to Austin’s challenge at the cost of positing a troubling form of semantic blindness. This is because, as we saw in the previous chapter, contextualism has to disrespect ordinary intuitions about the dialectical connections between some cross- context knowledge claims (and their metalinguistic variants). But the epistemic noncognitivist position that I have been sketching has a way out of this problem. The basic idea is that by
maintaining that knowledge claims express states of norm-acceptance rather than relativized descriptive beliefs, the epistemic noncognitivist like the ethical noncognitivist gains a second axis of possible opposition or agreement. That is, he can account for the intuition of cross- context dialectical connections between knowledge claims by pointing out that, on his view, knowledge claims can express pragmatically opposed or concurring states of norm-acceptance. Finding logically contradictory or identical descriptive beliefs is not the only way to account for dialectical connections.
More intuitively, it is easy to see how a speaker’s willingness to treat someone as entitled to her belief that p—i.e. acceptance of particular epistemic norms—might vary with the speaker’s changing circumstances. If I am just idly speculating about whether S knows that p and am not especially worried about possible defeaters, I will be perfectly willing to treat someone as entitled to her belief even if it is based on good but perhaps not terrific evidence. However, if my circumstances change and I need to rely on S as an informant about whether p, I may become worried about possible defeaters; then, if I suspect that the person doesn’t have terrific evidence for her belief, I will be less willing to treat her as entitled to the belief. In this way, the norm- expressivist can explain the intuitive variations in the attributability of knowledge across conversational contexts without positing the sort of semantic blindness about dialectical connections that the contextualist was forced to posit.
Of course, different epistemic norms might entitle beliefs for importantly different reasons. For example, we could imagine reliabilist norms which entitle beliefs formed by a reliable belief forming process and reason-based norms which entitle beliefs which are based on sufficient reasons. So, in some ways, this proposal is neutral with respect to traditional epistemological debates about the nature of knowledge. In fact, I suspect that part of the reason that such traditional debates have seemed so intractable is that metaepistemologists have illicitly assumed that knowledge claims are descriptive and thus taken it for granted that knowledge claims have
turn encourages the thought that there is some fact of the matter about whether reliabilists or internalists are right about the nature of the worldly relation of knowledge. However, if we give up this assumption in epistemology, as the expressivist tradition has urged us to do in ethics, we gain the resources to explain why the differences seem intractable, while neither resolving them nor dismissing their importance.94
Regarding the challenge of the paradox, the epistemic noncognitivist can mimic the contextualist resolution of the paradoxes. The core idea behind the contextualist’s strategy for resolving the AFC paradoxes was to treat the conclusion of the skeptical AFC arguments as true in the context in which the argument is presented; but this is consistent with the view that many ordinary knowledge attributions are true in the more quotidian contexts in which they are made. Whatever plausibility this response to the AFC-paradoxes gains for contextualism can be just as easily captured by the epistemic noncognitivist position I have been sketching. The strategy will be to transmute the contextualist’s notion of varying contextual standards for knowledge into the notion of varying states of epistemic norm-acceptance. That is to say that the epistemic noncognitivist will argue that the AFC arguments have the pull that they do insofar as they get us to accept more demanding epistemic norms than we typically do in ordinary contexts. But the resistance to their conclusion is due to the fact that we harbor acceptance of the less demanding standards.
Contextualism achieves its answer to the challenge of the paradox at the cost of positing semantic blindness to anyone who has been troubled by the AFC-paradoxes, which generates the worry that it is an ad hoc response to the paradoxes. Importantly, however, the epistemic noncognitivist, unlike the contextualist, treats the skeptical conclusion as in conflict with ordinary knowledge claims; and this explains why, in a sense, it is right to be troubled by the
94 For an independent argument that at least some epistemological discourse is partially nonfactual, which
is motivated by similar observations about the intractability of certain sorts of epistemological debates, see Field (1998).
paradox. To be sure, for the norm-expressivist, the relevant conflict is not (typically) a conflict between logically contradictory descriptive beliefs but rather a disagreement in which norms are accepted. A further advantage of this response to the AFC-paradoxes is that it can explain why some people find the paradoxes more troubling than others. Insofar as one is either a convinced skeptic or a convinced anti-skeptic, the contingencies of, respectively, ordinary epistemic discourse and skeptical arguments will not sway one’s acceptance of particular epistemic norms. But if one’s acceptance of particular epistemic norms is more flexible, then one will more easily be drawn into a situation where it seems as if some knowledge attribution is both true and false. This allows for a metaepistemological theory that is importantly neutral with regard to certain normative epistemological debates. For by making the meaning of knowledge claims depend on the states of norm-acceptance expressed by whoever makes them, the epistemic noncognitivist is not forced to take sides on who is right in a debate between skeptics, anti-skeptics, and those who cannot make up their minds.
So far, epistemic norm-expressivism appears to answer Austin’s challenge and the challenge of the paradox just as well as contextualism, and, since contextualism answered these challenges better than any of the various invariantist positions, this suggests that norm-expressivism is better than them with respect to the first two challenges. Moreover, epistemic norm- expressivism appears to meet the challenge of semantic blindness better than contextualism, and so it does not have the central drawback of that view.
We should recognize, however, that the norm-expressivist strategy for meeting the challenge of semantic blindness is not completely satisfactory. In effect, the epistemic norm-expressivist co-opts the contextualist’s answer to Austin’s challenge and the challenge of the paradox, but then he adds an extra noncognitive element—norm-acceptance—in order to avoid positing semantic blindness. I think this is a better package than contextualism, but an opponent could complain that this is an ad hoc maneuver. For it appears to generalize and provide a formula for