The Merits of Incoherence
2018; Wiley; Volume: 59; Issue: 1 Linguagem: Inglês
10.1111/phib.12118
ISSN2153-960X
Autores Tópico(s)Epistemology, Ethics, and Metaphysics
Resumo1. My main interest when doing epistemology is in the conditions, nature, and "logic" of a status or quality that folk language may have no unambiguous direct expression for. I can direct the attention of theorists to this status by calling it "prospective justification or warrant to be more confident" that something is the case. 2. I say "justification or warrant" because I make no subtle distinctions between these, as some other authors do. Another way I've sometimes captured this dimension of the quality is to say that it's epistemically more appropriate for you to be more confident. But with all of these expressions, we need to rely on examples to separate them from their intuitively close kin. I say "prospective justification," meaning thereby what some authors (including past versions of myself) call "propositional justification." I've come to find the latter phrase unfortunate, because it can suggest that what's being justified is (in the first place) a proposition, rather than an attitude like believing or suspending, or some kind of attitude dynamic, like becoming more confident. But that ought to be a substantive question, not one we hard-code an answer to in our shared vocabulary. (Also, I am sympathetic to the "attitude" answer rather than to the "proposition" answer.) Another thing the phrase "propositional justification" can suggest is that where the justification comes from is some set of propositions, those which constitute your evidence. Again, that is, a substantive commitment, that our shared vocabulary ought not to decide prematurely. (Again, this is a commitment I would myself resist.) I say "to be more confident" meaning to be more confident in some hypothesis than you should/would be in the absence of that justification. Later in this discussion we'll encounter an opportunity to distinguish between having more justification for confidence in P, and having (some, or sufficiently much) justification to be more confident. For the moment, I want to treat these indifferently; but I haven't found a vocabulary that sounds neutral between them. You'll notice that I talk about confidence rather than all-or-nothing doxastic stances like assenting or believing. It is controversial which (if either) of these kinds of doxastic stances we should think of as more fundamental. I have leanings; but I think we can side-step this controversy. Whichever of these kinds of stance is more fundamental, I assume it is feasible and can be productive to engage in theoretical inquiry about when one should be more confident, without also thereby inquiring into when it's appropriate for one to close inquiry or "settle" into an all-or-nothing doxastic stance.1 For ease of reference, let me abbreviate "prospective justification or warrant to be more confident" as PJC. 3. It's natural to divide the conditions that contribute to your having PJC for some hypothesis Q into two exclusive categories. (I don't know whether they are exhaustive.) Some of these conditions are the kinds of things we'd be willing to call "your justification for" Q, and which could make up "your grounds" for having some doxastic stance toward Q. (We're interested in these conditions even when you don't take any stance toward Q, but merely ought to; our question here is whether the conditions are such that, and then so related to you that, you could respond to them qua grounds.) Perhaps there are two questions here: whether the conditions merely could be your grounds, and whether you could reasonably so rely on them. But let's pass over that complexity for now. The other category are conditions that may play some role in making you justified, but that we only regard as enablers of your having that status, rather than "your justification" or the kind of thing you could sensibly base a doxastic stance on. By PJC I mean only to include things from the first category: potential grounds, rather than mere enablers. Epistemologists use the term "defeater" in several ways. Some use it to include considerations that work like negative enablers. This usage was especially prevalent in the Gettier literature, but is not confined to those discussions. On the other hand, many contemporary epistemologists, including myself, use "defeater" to mean a kind of potential ground. Defeaters come in several varieties. One straightfoward kind of defeat of your PJC for Q would be some PJC for not-Q, or any hypothesis that's clearly incompatible with Q. This can be called "opposing" or "rebutting" defeat, though we should not construe the words "rebut" or "defeat" here as success terms. The kind of defeat we're considering comes in degrees and can itself be defeated. A more complex and interesting variety of defeat is what theorists call "undermining" or "undercutting" defeat. Again, this can come in degrees and itself be defeated, so the labels should not be understood as success terms. I've argued that this category shouldn't be thought of as excluding the more straightforward one—there can be "mixed"/hybrid cases—and also that we shouldn't take it to be obvious that the two categories exhaust the space of defeaters.2 It is very tricky to define or explicate what makes something an undermining defeater. I will take it for granted that we all understand this category well enough to inquire into it, and can agree about some paradigm examples. One controversial issue we'll be considering below is whether higher-order PJC—that is, PJC for some (perhaps false) claim about the epistemic facts in general or about your own epistemic status—can be any sort of defeater for the corresponding first-order issues. For example, if you have some good first-order grounds to be confident that Q, and then acquire PJC for the (false) claim that you don't (and never did) have such grounds, does that on balance justify you in being less confident that Q? If higher-order PJC can have this kind of defeating effect, I will assume it counts as a kind of undermining defeat. This is a substantive assumption, though many others make it too. There may be important differences between this kind of undermining defeat, and more commonplace examples, such as evidence that your instruments/senses aren't reliable about Q.3 I don't mean to preclude such differences. I assume only that there are also important similarities, and that we can productively theorize about them to some extent as a single class. Undermining PJC may have a positive counterpart (e.g., evidence that your instruments/senses are more reliable than you thought). We could call this bolstering PJC. 4. I've defended an epistemology of perception I called "dogmatist." I wasn't the first to defend the kind of epistemology in question, and nor was I the first to use that name in a philosophical discussion. Though in doing so, I started a thread of using that name for more-or-less the kind of epistemology I defended. My use of "dogmatist" was inspired by its original application to Stoic epistemologists, who thought against their Skeptic contemporaries that it was possible for us to have reasonable beliefs. There are of course other usages of "dogmatist" and its cognates in philosophy: there's the pejorative use, and also a use to refer to the Kripke/Harman puzzle about whether evidence for Q entitles one to expect any future evidence against Q to be misleading.4 There's also Richard Jeffrey's use of "dogmatic" to refer to theories that prioritize all-or-nothing belief over degree-like doxastic attitudes. Let's ignore all these other usages, and confine our attention to the thread of using "dogmatist" for more-or-less the kind of epistemology I defended. A lot of variation in usage still remains, which it might be useful to sort out. Before I explain my understanding of "dogmatist," though, and how it differs from some others' usage, let's consider a neighboring terminological thread. This concerns the pair of terms "liberal" and "conservative." These terms also have usages in philosophy that are irrelevant to our concerns here. The term "conservative" even has another usage in epistemology, to mean a "negative coherence theory" that says attitudes count as being justified, or absolved of the need for justification, just by being had. The thread that I've participated in of using these terms in epistemology (I think I started this one, too) is different. I use the terms to describe exclusive but not exhaustive positions a theorist can hold about a chosen epistemic flaw or vulnerability. To illustrate, suppose Jessie has what is or at least seems to be PJC for the hypothesis H that she has hands. Now attend to considerations or hypothetical evidence such that, were Jessie to become so related to it that it became a potential ground for her, it would undermine the PJC she has for H. Call such considerations or evidence U. What epistemic relations does Jessie need to have toward U, as part of her having PJC for H? The liberal position says that Jessie can have PJC for H without needing any "antecedent" PJC against U. She doesn't need any independent reason to "rule out" U in advance. Nor does U (or its threatening content) need to in fact be false. But if Jessie does go on to learn or acquire U as evidence, or get PJC for it, that will undermine/defeat the PJC she otherwise, and until then, had for H. The conservative position, on the other hand, says that Jessie does need to have PJC against U, as part of having PJC for H in the first place. She may be able to rule out U because of some "default entitlement," rather than from any inquiry she engaged in. But without PJC against the potential underminer U, Jessie can't have PJC for H in the first place. Some forms of this view might add that U also has to in fact be false, for Jessie to have PJC for H. Perhaps she even has to know not-U, to have PJC for H. But I understand those as stronger forms of conservatism, not part of the label's meaning. These views are not exhaustive. A third position will say that U does have to in fact be false, for Jessie to have PJC for H, though she doesn't need antecedent PJC to believe that it's false. If U is "My senses are unreliable," then reliabilists about perception provide a good example of this third view. They say that Jessie will have (perceptual) PJC for H only if U is false, that is, only if her senses are in fact reliable. She doesn't need evidence that her senses are reliable. And even if her senses are reliable, evidence that they aren't (i.e., for U), will defeat the PJC she has for H. There may be yet further positions. When undermining evidence comes to light, some epistemologists are inclined to say that it shows that the subject never really had justification for H in the first place. I'm not sure whether this is best understood in terms of the third position I described, or as something new. As I understand these positions, they are not monolithic. One might be liberal about some epistemic vulnerabilities, conservative about others, and have the third position about yet others. I'm not sure that it's coherent to be liberal about all vulnerabilities. In my view, one kind of possible underminer for my PJC for H is the claim that I lack such PJC. Evidence that I lack such PJC would defeat (contribute toward defeating) that PJC, even if I in fact had it. In this case, surely the right attitude to take to the underminer is the third position. For me to have the PJC, this possible underminer really does need to be false. It's not enough for me to merely lack evidence to believe it's true. I'm not sure that it's coherent to be conservative about all vulnerabilities, either. But this is a position that some epistemologists are attracted to. 5. Now let's return to the term "dogmatism." I've sometimes defended an epistemology of perception that I applied that label to.5 At some point, though, I realized that the commitments that I was taking to be definitive of "dogmatism" were shared by a large family of views, where I had just been speaking in favor of one specific subspecies of the family. The specific view I advocated was distinguished by being internalist, concerning our PJC for observations about our perceptual environment, and giving a central explanatory role to the phenomonology of perceptual experience. Other views that differed in these respects could also be "dogmatists," as I came to understand that term. To give just two examples, one could also call oneself a dogmatist if one thought that PJC (or whatever epistemic status one was working with) required one's senses to in fact be reliable, or if one thought that PJC was only acquired when one genuinely perceived, not when one hallucinated. Around the same time I was advocating my version of "dogmatism," Michael Huemer was defending a view of much the same sort, that he called "Phenomenal Conservatism."6 That label applies more specifically to the subspecies of dogmatism that he and I are most sympathetic to, rather than to the whole family. Huemer also defended this view more broadly than just for perception. I am sympathetic to some generalizations (e.g., to memory and to math), but I am also open to there being other sources of justification, which he resists. We don't agree about all the arguments we each give for our views, but of course there is some overlap. Huemer and I weren't the first to defend dogmatist views. Roots of our views can be seen in the common sense philosophy of Reid and Moore, and the particularism of Chisholm.7 You can see views more recognizably like ours in Pollock's Knowledge and Justification (Princeton, 1974), and in a series of "modest foundationalisms" that were developed in the 1970s and 1980s. These views were clearly members of the broad dogmatist family, as I understand it, though many of them weren't Huemer's and my favored kind of dogmatism. Pollock, for example, gave a central role to a neo-Wittgensteinean theory of concepts, and others of our predecessors gave a central role to a non-propositional notion of acquaintance. Also, not all of our predecessors here were foundationalists about perception: some thought we had immediate but defeasible justification only about internal subject matters. I do believe in immediate but defeasible justification, but I don't count myself as a foundationalist, because that requires additional commitments about what other sorts of justification are or aren't possible, and about the global structure of justification. 6. In the field, not everybody uses the term "dogmatism," nor the terms "liberal" and "conservative," in the ways I've explained them here. Some use "dogmatist" to refer to the specific views Huemer and I advocated about perception. Some associate that term with specific commitments about the legitimacy of Moore-style proofs about the external world. I have myself argued that Moore-style proofs aren't guilty of all the vices they've been charged with (though they're no saints either).8 And I think there are natural motivational connections between dogmatism and the arguments I gave there. But I regard these connections as substantive, not part of the definition of "dogmatism." I'd call some views "dogmatist" that rejected my views about Moore-style proofs. But as I said, some other authors do understand "dogmatism" to include extra commitments about Moorean arguments. It's common also to see the term "liberal" used interchangeably with "dogmatist." What I would myself say is: being a dogmatist commits you to non-conservatism about every potential underminer. But one may take a liberal position wrt some underminers, and the third position wrt to others—in fact this may be the only coherent overall liberal view. If you're reading any given author who uses the terms "liberal," or "dogmatist," or some nearby term like "neo-Moorean," and you want to know specifically what they mean by it, you'll have to look closely at what that author specifically says and/or assumes. There seems to me to be too much variation in the details of different authors' usage for us to have a shared exact understanding of these labels. 7. In "Problems for credulism" (op. cit.), I proposed the label "credulism" for a family of views that was broader than dogmatism. (The motivating idea was that being credulous was somewhat like being dogmatic, though perhaps not so extreme.) In terms of the notions explained above, credulism can be identified with the view that we should be non-conservative about some potential underminers. A credulist is allowed to say we should be conservative about others. For example, one might plausibly say that a subject who has arithmetic justification for some result can have that justification be undermined by evidence that she's taken a bad-at-math-pill, without saying that any antecedent reason to think she hadn't taken such a pill needed to be part of her original justification. That would count as a credulist view of this species of justification. It permits one to affirm or to deny the further claim that any arithmetic justification is immediate. (Perhaps arithmetic justification always requires one to have antecedent PJC for the Peano Axioms or the Principle of Non-Contradiction.) Whereas only a few epistemologists endorse (i) the specific views Huemer and I advocated about the epistemology of perception, a much broader group endorse (ii) views that fit my broad understanding of "dogmatism" (including dogmatic accounts of justification other than perceptual), and even many more endorse (iii) views that fit my understanding of "credulism." Indeed, among theorists with explicit commitments, anti-credulists are in a small, polarized minority. Even BonJour when he was a coherentist was a credulist (and arguably even a dogmatist, about our justification for claims about what we believe). Curiously, most of the objections I've encountered to (i), or at any rate, the ones that my opponents seemed to be most moved by, really seemed to most fundamentally be challenges for (iii). My specific views (i) may well be wrong, but many of us together have to find some way to meet these challenges to credulism. This paper explores one such pressing challenge, and how we might respond to it. You may well find my response to this challenge a cure that's worse than the disease. 8. We'll get to the challenge, and my response to it, only in the closing Section 9 of this paper. In between, we need to introduce and coordinate our assumptions about a range of general issues about justification, undermining evidence, and incoherent attitudes. Consider Matthew, who has just proved some theorem T. Consider also Noma, who seems to herself in the same way to have proved a theorem T′. That is, Noma has all the superficial experiences of having proved T′. But in her case, let's suppose this is an illusion. She's made a subtle mistake. You are welcome to suppose that T′ is or isn't really provable; I require only that Noma isn't aware of a proof of it, but only seems to be. Perhaps she even has produced a proof, but her understanding of it is critically flawed in ways she's not aware of. So it merely feels to her like she's successfully grasping a proof of T′. In some such cases Noma may have inductive grounds for thinking that T′ probably has been proved. Perhaps whenever she seems to have proved something, it tends to be true. Or depending on her track record, maybe she'll have inductive grounds for believing the opposite! For the present discussion, though, I want to set any such inductive grounds aside. Let's suppose Noma lacks any inductive evidence about the track-record of her mathematical seemings. She just has the experience itself, of seeming to have proved and understood the proof of T′. I'll want to refer to some versions of these stories where that is all the evidence Matthew and Noma have. Call the subjects in those scenarios Matthew-0 and Noma-0. I'll also want to refer to versions of these stories where Matthew and Noma get additional, "higher-order" evidence, telling them that they aren't justified in their conclusions. For example, perhaps their brilliant mathematical aunt looks over their work and tells them it isn't a proof. It's important for these versions of the stories that the new evidence really does justify Matthew and Noma in doubting that they had a proof, and doubting they were justified in believing their respective "theorems." In Matthew's case, though, those claims are false. He did have a proof, and before his aunt showed up, at any rate, he was justified in believing his theorem. We can call Matthew's new evidence misleading because it's evidence for false claims. Nonetheless it does give him justification for those claims. Call the subject in that scenario Matthew-M. ("-M" for "misleading (higher-order, defeating) evidence.") In Noma's case, the aunt as I've described her may be telling Noma something true. That is, she may be right that Noma lacks a proof. But we can imagine versions of the story where Noma's aunt is still misleading her, because she's just making up a criticism, and doesn't really believe she lacks a proof; and also versions where Noma's aunt is in fact pointing out to her exactly what the mistake in her initial reasoning was. Let's call these subjects Noma-M and Noma-Revealed. I'll also want to refer to versions of these stories where Matthew and Noma don't have any higher-order evidence, but they do have negative higher-order attitudes—attitudes they (arguably) ought to have formed in response to the defeating evidence described a moment ago. They may have these negative higher-order attitudes because of a crippling insecurity. On some views, they may have them in response to illusions of acquiring the -M evidence. But it's important to these versions that Matthew and Noma merely believe—while lacking justification for believing—that they don't have a proof and aren't justified in accepting their "theorems." In other versions they may merely doubt this, or be agnostic about it. Call all such subjects Matthew-A and Noma-A. ("-A" for "akratic" or "agnostic.") It will make discussion easiest if we assume that in the -M cases, Matthew and Noma do also form the higher-order attitudes described in the -A cases. (Some of our discussion should carry over to the cases where they don't.) It will also make discussion easiest if we assume that in all the cases, Matthew continues to believe T on the basis of his (genuine) proof, and Noma continues to believe T′ on the basis of her (illusory, or ill-understood) proof. In some of the versions, then, they are believing in ways that their higher-order critical beliefs tell them they shouldn't. (Some of our discussion should carry over to cases where Matthew and Noma don't do this.) Here is a table of these possibilities: I label Matthew-0 "uninteresting" because it seems straightforward what our epistemological theories should say about him. He has and understands a proof; so presumably he's justified in believing on that basis the theorem he's thereby proved. Of course there will be different explanations of the nature of that justification; but what I've said so far won't be contested. I label Noma-Revealed "uninteresting" because it also seems straightforward what our epistemological theories should say about her. She seemed to grasp a proof, but an informant she has reason to trust has explained to her what her mistake was; so presumably she's not justified in continuing to believe her "theorem." Again, there will be different explanations of how and why that works. But what I've said so far won't be contested. The remaining cases: the -M and -A cases, and Noma-0, are more interesting and provoke a variety of theoretical positions. S1. Noma-0 is just as justified in believing her "theorem" as Matthew-0 is in believing his. S2. All the agents with higher-order attitudes (whether based on evidence or not) that say they are in a deluded, Noma-like condition, are unjustified in believing their "theorems." O1. Only Matthew (-0, -M, and -A) has any justification for his "theorem"; Noma-0 has none. (Except perhaps for inductive justification, which we're here ignoring.) O2. What's important for Matthew's justification to believe T is just whether he does in fact grasp a proof. So he's entitled to ignore the higher-order evidence or attitudes that mislead by saying he doesn't. At least, he can ignore them when it comes to the question of T's truth. I'm not happy with, or willing to straightforwardly accept, any of these four claims. I think the truth is somewhere in the middle of this loose spectrum, and will require some care to articulate (much less to convincingly establish). Rather than "objective," sometimes that end of the spectrum is instead called "externalist" (and the other end "internalist"). This can be confusing, since there's no direct, definitional connection between the kind of "objectivity" described here and externalism in the sense associated with Goldman, nor the sense associated with Williamson. Sometimes all the views not at the "objective" (or "externalist") end of the spectrum are called "subjective" (or "internalist"). Other times all the views not at the "subjective" end are called "objective." These terminological practices are also confusing. I have another, specific reason for discomfort with the term "subjective," as some hear this to mean that it's the subject's perspective that determines her normative relation toward a proposition Q. But on many views that oppose the most "objective" end of the spectrum, the subject's higher-order evidence and attitudes needn't determine her normative relation toward Q all by themselves. They may contribute to making the subject stand in some such relation, but they won't trump other contributing factors, and arguably there can be other contributors that are unrecognized by the "objectivist" than a subject's higher-order evidence and attitudes. For all these reasons, I dislike and prefer to avoid the qualifiers "objective" and "subjective" (or "external" and "internal"). These terms do more harm than good in these discussions. But it is helpful to have in mind the loose arrangement of views I've used those terms to introduce. With respect to S1 versus O1: I'm inclined to deny that Noma-0 has nothing by way of justification. But I'm also inclined to doubt that she's epistemically on equal footing with Matthew-0. There seems to be some positive epistemic relation she stands in to T′, but something better about the relation that Matthew stands in to T. (Even if T′ and T do both happen to be theorems.) I wish I understood this more, and were in a position to say more about it (or even knew what were the right questions to ask, to get us moving forward). But I don't. Nonetheless, this is how I'm inclined to regard these subjects, and in discussing these cases with others, I've found that many others, perhaps most others, are also so inclined. But there is no consensus. The rest of our discussion will focus on what to say about the -M and -A cases, for both Matthew and Noma. Q1. What are the normative effects of mere (unjustified) attitudes? This work appears under labels like "wide-scope oughts" and "structural normativity," and one of its central contributors is John Broome, but there are also many other labels and contributors. Q2. What are the normative effects of higher-order information? In epistemology, one of the central contributors to this work is David Christensen, but there are also many others. These first two questions interact with each other, when we consider cases like Matthew-A and Noma-A who have mere (unjustified) higher-order beliefs, rather than higher-order evidence. Also, much of the past decade's work on "the epistemology of disagreement" has engaged with aspects of both Q1 and Q2. Q3. Can it ever be reasonable (or justified or rational) to be inconsistent, akratic, or so on? Later, I'll introduce an general umbrella term "incoherent," and will restate Q3 as: Can it ever be reasonable to be "incoherent"? Q4. Are "normative tragedies" or "dilemmas" or "rational toxicity"9 possible? that is, cases where the subject has no option she can reasonably take? After developing my combination of answers to these questions, I'll talk through what it tells us to think about our -M and -A cases. Ask a dozen different ethicists or epistemologists their views on the four issues listed, and nowadays you're likely to hear a dozen (or more) different combined accounts. I cannot hope to survey all the candidates, or to do more than begin to motivate the combination I find most workable. But as we walk through these questions, I'll gesture toward some other popular accounts besides the ones I'm recommending. Let's begin with questions about the normative effects of mere (unjustified) beliefs. Can such beliefs justify us in believing their obvious inferential consequences? One extreme view (from the "objective" end of Section 2's spectrum) will say these beliefs have no normative effect at all. Only what justification (or: what justified beliefs) you have matters. The other, "subjective" end of the spectrum will say that justification is just a matter of the downward dynamics from your subjective perspective, that is, the beliefs you start with. So "mere" beliefs aren't in themselves handicapped as justifiers. If you believe P, and P obviously entails Q, then you ought to believe Q (or: then believe Q!) If you believe you ought to believe Q, then you ought to believe Q (or: then believe Q!) This debate also engages with several principles of practical reason, but I'll confine our focus to epistemology here. We'll return to principles like the second in Section 5, below. Observe that in the antecedents of these principles, mental states are specified that we're not assuming to be justified. The "narrow" interpretation of the principles sees them as having the form: A → Ought B, and so merely havin
Referência(s)