Abstract
The paper outlines a view of normativity that combines elements of relativism and expressivism, and applies it to normative concepts in epistemology. The result is a kind of epistemological anti-realism, which denies that epistemic norms can be (in any straightforward sense) correct or incorrect; it does allow some to be better than others, but takes this to be goal-relative and is skeptical of the existence of best norms. It discusses the circularity that arises from the fact that we need to use epistemic norms to gather the facts with which to evaluate epistemic norms; relatedly, it discusses how epistemic norms can rationally evolve. It concludes with some discussion of the impact of this view on “ground level” epistemology.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
My first published attempt in this direction (aside from very sketchy remarks near the end of Field 1982) was a discussion of nonfactualism about normativity generally, in Sections 3 and 4 of Field 1994 (republished with significant corrections in Field 2001), which was much influenced by Gibbard 1990. The application to epistemology was discussed in Field 1998b, 2000, 2005. The view is at least somewhat similar in spirit to views recently expressed by Max Kölbel 2004 and John MacFarlane 2005, and I have benefited from thinking about these works, as well as from discussions with numerous students and colleagues including Sinan Dogramaci, Gary Ebbs, Andy Egan, David Enoch, Melis Erdur, Dana Evan, Paul Horwich, Matt Kotzen, Maria Lasonen-Aarnio, Jim Pryor, Karl Schafer, Stephen Schiffer, Ted Sider, Sharon Street, David Velleman, Lisa Warenski, Tim Williamson, and Crispin Wright. Paul Boghossian and Stewart Cohen deserve special mention: Boghossian’s persistent skepticism about my views has led to a considerable sharpening of them, and Cohen’s extensive correspondence about the near-final draft led to some substantial improvements. Andy Egan and Seth Yalcin have pointed out connections between some of the ideas in the paper and recent work on epistemic modals, by them and others, but I’ve decided not to expand an already long paper to pursue this.
While distinctions between these terms are possible, I don’t believe there is any standard such distinction. Perhaps the main difference is that ‘justified’ seems ambiguous: calling a belief justified often means that it’s reasonable, but occasionally means that there is something which serves as a justification for it; illicitly sliding between these two readings is a common source of epistemological chicanery. For this and other reasons, I prefer ‘reasonable’, and will generally use it in the rest of the paper.
For instance, (i) the usual metaphysical (Humean) worry, that there seems no room for “straightforward normative facts” on a naturalistic world-view; (ii) the associated epistemological (Benacerraf-style) worry that access to them is impossible (which is compounded by the fact that there is substantially greater disagreement about normative matters than about mathematical); (iii) the worry that the relation to norms is not only non-naturalistic, but “queer” in the sense that it’s supposed to somehow motivate one to reason in a certain way (Another worry related to (iii) will be mentioned near the end of Section 8).
I take no stand on whether the location is needed to determine which proposition is expressed, or is instead used as a parameter in terms of which the proposition is evaluated. This distinction (between “indexicality” and “non-indexical contextualism”, in current terminology) strikes me as more a matter of legislation on the use of ‘proposition’ than a matter of substance.
Indeed, there typically isn’t an intended norm, or even a small set of norms that are candidates for being intended: one’s normative views tend to be far looser than would be required for this.
The focus of the paper is on justification or reasonableness rather than knowledge, but I’m inclined toward some kind of evaluativism for knowledge too. One approach to knowledge takes it as involving justification to a sufficiently high degree that is undefeated. It seems to be widely accepted today that the required degree is somehow relative—usually this is taken as some sort of contextual relativity, but see Chrisman 2007 for a more expressivist variant similar in spirit to the kind of expressivism advocated in this paper. Most of this recent literature however involves relativism or expressivism only about the degree of justification required for knowledge; whereas I’m applying it to the standards of justification itself.
Of course, it’s controversial that knowledge requires justification at all; and if it doesn’t, the case of knowledge would require a quite separate discussion. But whichever way one goes on that, it seems to me that knowledge is a normative concept, and that straight naturalistic accounts (e.g. in terms of reliability) simply don’t do justice to it; and I think that something in the broad ballpark of what I say here would still apply. But that’s too big a matter to discuss here.
Like Dworkin (1996) I am skeptical that a hard and fast line can be drawn between ground-level normative claims (“internal claims”) and meta-normative claims (“external claims”): I’m inclined to agree that any sort of anti-realism about normativity is bound to affect ground level practice. It needn’t affect it by actually denying the realist’s claims: if a realist wants to say that it is a straightforward objective fact, out there, independent of us, that one shouldn’t kick dogs for the fun of it (cf. Gibbard 2003, p. 186), the appropriate reaction of the anti-realist isn’t to take a stand on the claim but to question the rhetoric and give a story on which the rhetoric is at the very least highly one-sided and in most contexts thoroughly misleading. (In the cited passage Gibbard advocates an internal adequacy thesis that seems to accept the hard and fast line, and the quasi-realist program. However, he takes the judgement about dogs to be ambiguous between an internal and external reading, and says that the rhetoric is “sumptuous” on the internal, so I’m not sure the extent of real disagreement. Still, adopting the aim of sounding like the realist seems to me to distort the enterprise.).
Here and in what follows, I mean by ‘time order’ linear time order: I’m not considering the non-frame-relative partial order given by the light cone structure.
If we do say there has been a change of meaning, say in ‘simultaneous’, what are we to say that the term prior to Einstein “meant”? If it “meant” a relation obeying all the properties that Newton assumed, then since there is no such relation, virtually everything speakers said using the word was false. If it “meant” some two-place relation that is relativistically kosher, then some very surprising claims come out true: for instance, if it “meant” something that coincides with space-like separation, then in Newton’s mouth “There is an event simultaneous both with the birth of Thales and with the birth of Galileo” would have been true. Given this, it may be best to interpret Newton as having used the word ‘simultaneous’ in a way that actually was frame-relative, though he didn’t know it. (Similarly for earlier uses of ‘parallel’.)
Even if we resist this view about Newton, we may want to maintain it for those today who are ignorant of Special Relativity and wouldn’t understand explicitly relativized simultaneity claims. If, as is usually assumed, the meaning of ‘arthritis’ for the lay person is determined by what the experts in his community mean, then it’s natural to suppose that in the mouths of lay people ‘simultaneous’ and ‘parallel’ have a hidden relativity that they don’t know about.
This is something one often hears in conversation, and there are at least strong implicatures of it in print. For instance, Boghossian (2001, pp. 22, 23) takes relativist and other non-factualist views to imply that no norm is “more correct than any other”. Read literally I can’t object: since talk of correctness (construed “thickly”: see Section 7) doesn’t apply at all, then it applies equally. But the formulation certainly seems to suggest that the view has it that the norms are equally good; and indeed, Boghossian goes on to say that “there is nothing that epistemically privileges one set of epistemic principles over another” (p. 23). [Boghossian informs me that he regards this last quote as careless and not reflective of his views. In his more recent work he continues to say that according to non-factualism, all norms are “equally correct” (e.g. 2006a, pp. 62, 73) and “equally valid” (p. 2), but is careful not to say that they are equally good. He does however give some arguments which in my opinion draw their plausibility from a natural tendency to associate “equally correct” with “equally good”.].
For a familiar reason, we can’t regard incomparabilities as ties: A can be incomparable to B and also incomparable to C, even though C is clearly better than B; if we were to take incomparability to be a kind of equal goodness, we’d induce intransitivity into the better than relation.
In doing this I am influenced by the opening chapter of Gibbard 2003, which takes norms of rationality to be plans. Among the advantages of the non-propositional choice is flexibility: it makes it easier for, e.g., both the policies one follows and the policies one approves of to get into the discussion.
As Boghossian 2006b emphasizes, there are norms of explicit permission as well as of obligation or prohibition: a logically omniscient being could read the permissions from the obligations or prohibitions (p is permitted if it is not prohibited, i.e. if not-p is not required), but those of us who aren’t logically omniscient can’t compute what does and doesn’t follow from our explicit obligations, so we need a system with both explicit obligations and explicit permissions. (This introduces dangers of conflict: the system could both explicitly permit p and implicitly prohibit p, thus being practically inconsistent. But such is life.) One way to understand norms of explicit permission is as explicit rejection of policies (so that a norm isn’t really a policy, but the acceptance or rejection of a policy). Alternatively, the notion of policy could be broadened: say, as a partial function that assigns to certain pairs of input conditions and possible actions either ‘+’ (for explicit permission) or ‘−’ (for explicit prohibition); when there is no mark, permissions and prohibitions (and obligations, i.e. prohibitions of the opposite) can still be implicit in the other explicit permissions and prohibitions, but aren’t explicit. For simplicity however, I will ignore explicit norms of permission in what follows.
If you think that compatibility is itself normative, you can replace it by compatibility in a given system; this can be explained in syntactic or model-theoretic terms, which certainly aren’t normative.
I put ‘intend them as’ in quotes because there is some issue about how to understand it. Certainly in a typical utterances one does not explicitly think about the relativity in ‘simultaneous’, ‘prior to’, ‘energy’, ‘momentum’, ‘parallel to’ and all the other terms one knows involve some sort of relativity. Perhaps the best thing to say isn’t that one “intends” the relativity, but that one has a standing view that when one makes such utterances or thinks such thoughts, a relativized interpretation is appropriate. But I’ll continue to speak in the simpler way.
I’ll take a “possible world” to include the spatio-temporal location of the agent in the world: that is, it is to be what Lewis 1986 calls a centered possible world.
Of course, the worlds at which n declares ‘You should refrain from believing p’ false (not true) will not typically be the same as the ones in which it declares ‘You should believe p’ true: it will declare both false if it neither prohibits nor demands belief in p, and it will declare both true if it both prohibits and demands belief in p. (I haven’t required that norms be consistent, much less that they be “consistent relative to the facts of each possible world”.).
I don’t mean to take a stand on how easy it is to find statements in English with no evaluative elements: it’s compatible with what I say here that most or even all apparently descriptive predicates are to some extent “thick” with evaluation. I do hold, with Blackburn and many others, that it is important to disentangle the descriptive and evaluative elements in “thick” predicates: indeed, the excellent discussion of ‘cute’ in Sect. 4.4 of Blackburn 1998 shows that failing to do so undermines the possibility of normative critique of entrenched views.
I’m being a little loose here, because a person can be committed to many policies at the same time; if we want to define a unique notion of degree of belief for evaluative claims on this model, we need to single out one of them. This will presumably involve (i) figuring out which one or ones are fundamental enough to take as contributing to the desired notion of degree of belief, and (ii) if there is more than one that is fundamental enough, moving to the “union” of them, i.e. the policy that contains the prohibitions in each of them. (Of course, the union may be inconsistent even when the individual members are consistent, but I’ve observed in note 17 that the model under discussion does not require that policies be consistent.) It is also possible to let the policies that count as “fundamental enough” be a matter of context, in which case the notion of impure degree of belief will be as well.
One could define variant notions of impure degree of belief by using, say, the norm an agent uses in evaluating herself rather than the norm she is committed to; but I think that the resultant notions are less natural and less likely to be confused with a kind of pure belief.
Actually it’s probably better to mark the distinction between pure and impure slightly differently: instead of regarding the degree of belief in “If he’s justified in believing that q then he’s justified in believing that q” (or the conjunction of this and some straightforwardly factual claim) as impure because it is a degree of belief in a claim that contains an evaluative term, we can regard it as pure because the claim in question is norm-insensitive—the norm drops out of the determination of the degree of belief. But I won’t bother to distinguish between the evaluative and the norm-sensitive in what follows.
Indeed, some terms thought of simply as having contextual sensitivity do behave in some ways like the way I’m claiming normative terms behave. This is especially clear in vague terms like ‘rich’: while they admit of standard contextual relativity, there also seems to be a normative element as to where one places the border. (“You think having $5,000,000 makes you rich? What about having $100,000,000—now that’s rich.”) See Egan forthcoming.
There is a similar issue about ‘property’. Like ‘true’, this can be given a pleonastic use, on which any meaningful predicate without contextual relativity corresponds to a property. Given this, I wouldn’t want to say that normative predicates don’t express properties; I say instead that they don’t express straightforward properties.
The point made here is influenced by Dworkin (1996) and Street forthcoming; both say that the issue of realism is a normative doctrine. (I think they, or at least Dworkin, means that it is solely a normative doctrine: that there is no metaphysical issue, only the normative. This I do not accept.) I’m also influenced by conversations with Melis Erdur.
I don’t believe the point is adequately answered by noting that an erstwhile realist who discovered there to be no straightforward normative facts would almost certainly continue with those moral preferences, and give up her view that she must base moral preferences on beliefs about such facts. For it’s still the case that while she’s a realist she has those conditional preferences; and those conditional preferences strike me as morally objectionable.
Indeed, not just inconsistencies: also conditional inconsistencies, i.e. giving inconsistent recommendations on specific factual assumptions.
It would be somewhat more realistic to generalize from ascribing to each agent single measures μ and ν (on the worlds and precise norms, respectively) to ascribing a pair of (non-empty) sets of μ-measures and ν-measures, or a (non-empty) set of pairs of μ- and ν-measures. But psychological realism is better pursued by the means suggested in Section 11.
Here I have been greatly helped by both Stewart Cohen and Gary Ebbs.
Since the distinction is pragmatic rather than semantic, I’m not sure that there’s much point in thinking in terms of two kinds of parameter: we can instead think of one kind of parameter that can be given two different pragmatic uses. Indeed, the latter is probably preferable, for two reasons. First, the two-kinds-of-parameter picture encourages the view that the pragmatic distinction is quite hard-and-fast, and that is probably unrealistic. Second, for anyone who isn’t a possible-worlds realist, there is no real distinction between the two uses of a world-parameter: in MacFarlane’s terminology, the “world of assessment” and the “world of use” must be the same (there’s only the actual world); or in the alternative terminology suggested immediately below, truth and objective correctness coincide. So in the case of worlds (possible worlds realism aside), the two-kinds-of-parameter view makes for a distinction without a difference.
There is an ambiguity in ‘non-indexical contextualism’. The usual official explanation is semantic, but this semantics can have two very different pragmatics associated with it, and I think that MacFarlane and most others who use the term associate with it the pragmatics on which normative utterances are taken to be objectively correct iff they are true relative to the utterer’s norms. As long as one recognizes the alternative pragmatics, which employs no such notion of objective correctness but only a norm-sensitive truth predicate, then it makes no difference whether one calls this a variety of non-indexical contextualism or reserves the term ‘non-indexical contextualism’ for views with the usual pragmatics.
In saying that the assessor-relativism accords with the norm-relative truth conditions, I don’t mean to imply that one could explain the difference between assessor-relativism and contextual relativism in terms of truth conditions. That is certainly false. It is central to the transparent notion of truth that I’ve been assuming that meaning can’t be explained in terms of truth conditions; meaning “comes first”, and truth conditions reflect it. The distinction between assessor-relativity and contextual-relativity is, as I’ve said, a pragmatic distinction, having to do with their different roles as regards the social phenomenon of disagreement.
Norms can include preferences as well as policies, but these are different kinds of norms. The assumption that N is the highest-level policy-norm guiding S’s behavior does not preclude that S evaluates it, but it does preclude that S has a norm to change N in light of such an evaluation. It is even conceivable to retain a policy-norm despite a highest level evaluation that one shouldn’t: consider the theoretical skeptics who (apparently) have norms of evaluation that disvalue the policy-norms that they employ. This is highly undesirable, and casts serious doubt on the utility of the norm of evaluation, but it is not incoherent.
Such as the conditional whose consequent is that there are continuous functions mapping the unit interval onto the unit square and whose antecedent consists of the axioms of set theory plus the set-theoretic definitions of the notions appearing in the consequent.
I take this example to be normative: the issue is about the rules by which to reason. (See Field 2009).
Well, we could stipulate that the processing of the representations shouldn’t count as literally rule-governed since it is not represented; but this stipulation seems insufficiently motivated. It seems reasonable to regard someone as “following” a rule when (i) the person’s behavior by and large accords with the rule, and there is reason to expect that this would continue under a decent range of other circumstances; and (ii) the person tends to positively assess behavior that accords with the rule and to negatively assess behavior that violates the rule. (In the case of epistemic rules, the “behavior” is of course the formation, retention, or revision of beliefs.) This is vague, which reinforces the point next to be made, about the considerable indeterminacy involved in ascribing epistemic or other rules to a person.
I don’t claim that anyone who believes in metaphysical justification (“justificatory fluid”) needs to think that there is a single overall level of fluid that matters, as opposed to levels at seventeen different places (or the amounts of seventeen different fluids). Still, I think that the belief in metaphysical justification does tend to go with belief in a single standard, or at least a most privileged one, and so I have taken the “dipstick model” to include this additional commitment.
I don’t mean to suggest that it is only the evaluativist who could see the argument as erroneous on this ground.
Hilary Putnam’s suggestion that quantum-mechanical results could be seen as indirect evidence against the distributive law is no counterexample, for two reasons. The main reason is that his claim that we could handle problems in quantum mechanics by giving up the distributive law was woefully under-supported. But in addition, he gave no support at all for his claim that the role of the quantum-mechanical facts would be to serve as evidence against the distributive law, rather than as an empirical trigger to motivate re-thinking whether there was a conceptual case for the distributive law. Perhaps the role of quantum mechanical evidence would be like the role of the discovery that there is no Santa Claus might have in the development of a free logic.
If ‘platonism’ is defined simply as the belief that there are mathematical entities that aren’t physical, mental, etc., then I don’t think that the Benacerraf argument applies to all forms of platonism: see Field (1998a). But it may be unnatural to regard the positions it doesn’t apply to as forms of platonism.
References
Benacerraf, P. (1973). Mathematical truth. The Journal of Philosophy, 70, 661–679. doi:10.2307/2025075.
Blackburn, S. (1993). Essays in quasi-realism. Oxford: Oxford University.
Blackburn, S. (1998). Ruling passions: A theory of practical reasoning. Oxford: Oxford University.
Boghossian, P. (2001). How are objective epistemic reasons possible? Philosophical Studies, 106, 1–40. doi:10.1023/A:1013141719930.
Boghossian, P. (2006a). Fear of knowledge. Oxford: Oxford University.
Boghossian, P. (2006b). What is relativism? In P. Greenough & M. Lynch (Eds.), Truth and realism. Oxford: Oxford University.
Chrisman, M. (2007). From epistemic contextualism to epistemic expressivism. Philosophical Studies, 135, 225–254. doi:10.1007/s11098-005-2012-3.
Dreier, J. (2004). Meta-ethics and the problem of creeping minimalism. Philosophical Perspectives, 18, 23–44.
Dworkin, R. (1996). Objectivity and truth: You’d better believe it. Philosophy & Public Affairs, 25, 87–139. doi:10.1111/j.1088-4963.1996.tb00036.x.
Egan, Andy forthcoming. “De gustibus non disputandum est (at least not always)”, In R. Feldman & T. Warfield (Eds.), Disagreement (Oxford: Oxford University).
Field, H. (1982). Realism and relativism. The Journal of Philosophy, 79, 553–567. doi:10.2307/2026317.
Field, H. (1994). Disquotational truth and factually defective discourse. Philosphical Review, 103: 405–452; reprinted with significant corrections in Field 2001.
Field, H. (1998a). Mathematical objectivity and mathematical objects. In S. Laurence & C. Macdonald (Eds.), Contemporary readings in the foundations of metaphysics. Oxford: Wiley-Blackwell.
Field, H. (1998b). Epistemological nonfactualism and the a prioricity of logic. Philosophical Studies, 92, 1–24. doi:10.1023/A:1017147216094.
Field, H. (2000). A priority as an evaluative notion. In P. Boghossian & C. Peacocke (Eds.), New essays on the a priori. Oxford: Oxford University.
Field, H. (2001). Truth and the Absence of Fact. Oxford: Oxford University.
Field, H. (2005). Recent debates about the a priori. In T. Gendler & J. Hawthorne (Eds.), Oxford Studies in Epistemology 1. Oxford: Oxford University Press.
Field, H. (June, 2009). “What is the normative role of logic?”. Proceedings of the Aristotelian Society Supplement, 83.
Gibbard, A. (1990). Wise choices, apt feelings. Cambridge, MA (London): Harvard University.
Gibbard, A. (2003). Thinking how to live. Cambridge, MA (London): Harvard University.
Gibbard, A. (2008). Rational credence and the value of truth. In T. Gendler & J. Hawthorne (Eds.), Oxford Studies in Epistemology 3. Oxford: Oxford University Press.
Harman, G. (1996). Moral relativism. In G. Harman & J. Thomson (Eds.), Moral relativism and moral objectivity. Oxford: Blackwell.
Kölbel, M. (2004). Indexical relativism vs. genuine relativism. International Journal of Philosophical Studies, 12, 297–313. doi:10.1080/0967255042000243966.
Lewis, D. (1971). Immodest inductive methods. Philosophy of Science, 38, 54–63. doi:10.1086/288339.
Lewis, D. (1974). Spielman and Lewis on inductive immodesty. Philosophy of Science, 41, 84–85. doi:10.1086/288571.
Lewis, D. (1986). On the plurality of worlds. Oxford: Blackwell.
MacFarlane, J. (2005). Making sense of relative truth. Proceedings of the Aristotelian Society, 105, 321–339. doi:10.1111/j.1467-9264.2005.00178.x.
MacFarlane, J. (2007). Non-indexical contextualism. Synthese. doi:10.1007/s11229-007-9286-2.
Pollock, J., & Cruz, J. (1999). Contemporary theories of knowledge. Lanham, MD: Rowman and Little.
Street, Sharon forthcoming. Objectivity and truth: you’d better rethink it.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Field, H. Epistemology without metaphysics. Philos Stud 143, 249–290 (2009). https://doi.org/10.1007/s11098-009-9338-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11098-009-9338-1