The Dogmatism Puzzle

Abstract

According to the Dogmatism Puzzle, knowledge breeds dogmatism: if a subject knows a proposition h, then she is justified in disregarding any future evidence against h, for she knows that such evidence is misleading. The standard, widely accepted, solution to the puzzle appeals to the defeasibility of knowledge. I argue that the defeat solution leaves intact a residual dogmatist puzzle. Solving this puzzle requires proponents of defeat to deny a plausible principle that the original puzzle relies on called Entitlement, a principle stating roughly that knowing that a piece of evidence is misleading entitles one to disregard it. The plausibility of Entitlement should cast doubt not only on the defeat solution, but on an assumption that has often been taken for granted: the falsity of the dogmatist conclusion of the original puzzle. I conclude that we face a dilemma between giving up Entitlement and living with dogmatism.

Keywords:

• dogmatism puzzle
• knowledge
• defeat
• misleading evidence

Notes

¹ The paradox was first presented by Kripke [‘On Two Paradoxes of Knowledge’, unpublished lecture delivered to the Cambridge Moral Sciences Club].

² I am making as few assumptions as possible about what evidence is. I leave open, for instance, whether evidence consists of propositions, mental states, or both.

³ ‘Seeming to tell against h’ is an unhappy expression for Harman to be using. If evidence that tells against h is just evidence against h, then seeming to tell against h is seeming to be evidence against h. Reading ‘misleading evidence’ in the argument for dogmatism as ‘evidence that seems to be evidence against a truth’ would lead to an uncharitable reconstruction of the argument. As will become clear, the argument relies on the assumption that a subject is entitled to disregard evidence she knows to be misleading. But the assumption that a subject is entitled to disregard evidence that she knows seems to be misleading is far less plausible.

⁴ For example, assume that I start out thinking that a coin is fair, and it is in fact fair. I flip the coin, and the first toss comes up heads. This slightly raises the probability that the coin is biased in favour of heads. Therefore, it is (weak) evidence against a truth, namely, the proposition that the coin is not biased in favour of heads. Similarly, had the coin come up tails instead of heads, I would have acquired some evidence against a truth.

⁵ It might be helpful to distinguish between misleading evidence against h, on the one hand, and misleading evidence regarding h, on the other. There are two ways for E to be misleading regarding h: either h is true and E is evidence against h, or h is false and E is evidence for h. Now, one might think that something even more general than Entitlement is true: to be justified in disregarding E as it bears on h, all that a subject needs to know is that E is misleading evidence regarding h. This won't make a difference for the purposes of the Dogmatism Puzzle, since the puzzle assumes that one knows h, and on this basis comes to know the stronger claim that a piece of evidence E is misleading evidence against h. Note also that nothing I say turns on a difference between epistemic entitlement and epistemic justification.

⁶ I am aware of only one person who explicitly rejects it, Michael Veber [2004]. I discuss his argument below.

⁷ Now, there is a class of worries having to do with the second premise. One might argue either that (2) has a necessarily false antecedent and hence, is trivially true for reasons not having to do with knowledge defeat (one might try to do this by appeal to contextualist or subject-sensitive invariantist considerations), or that it is false because of closure failures. Though I cannot argue for this here, on closer inspection such attempts to block the argument do not seem particularly convincing—and proponents of the defeat solution ought to sympathize with me here.

⁸ A solution along these lines has been defended at least by Harman [1973], Ginet [1980], Sorensen [1988], Conee [2004b], Hawthorne [2004: 73], and Kelly [2008].

⁹ Someone might object that we shouldn't assume a philosophical account to take the form of necessary and sufficient conditions, for an account could take a disjunctive form: for instance, perhaps there are a certain number of ways in which one might disregard a piece of evidence, and the right account of what it is to disregard evidence takes the form of a disjunctive condition (cf. Williamson's [2000] view that knowledge is the most general factive mental state). But first, I see no reason to think that the right account of disregarding would take such a form. And second, what I will say should make it plausible that no disjunction of the conditions I will reject as being necessary would constitute an adequate account of disregarding evidence.

¹⁰ Veber [2004] seems to endorse this idea. He equates disregarding a piece of evidence E with not ‘getting’, i.e. accepting, E in the first place.

¹¹ Having E as evidence requires believing E on numerous views of evidence, whether a subject's evidence is seen as consisting of all the propositions she believes at a time, the contents of certain foundational beliefs, her justified beliefs, or the beliefs that count as knowledge. Also, Bayesian views on which updating occurs by conditionalization require that evidence-propositions are assigned a credence of 1. If one thinks that assigning a credence of 1 to a proposition entails believing it, then evidence-propositions will have to be believed by the subject.

¹² Thanks to an anonymous referee for suggesting this modification to (iii).

¹³ Several authors writing on Harman's paradox more or less explicitly assume that disregarding evidence regarding h would amount to not adjusting one's belief in h, or continuing to be equally certain of h as before. For instance, Sorell [1981: 563] talks about not doubting one's judgment, or continuing to be certain. Ginet's [1980] idea is that ignoring E as it bears on h is treating it as ‘no evidence for me against h, as not weakening the case I have for believing confidently that h is true’.

¹⁴ Though (v) refers to credences, I want to leave open the possibility of non-credence-like fine-grained belief-states (perhaps fine-grained states that don't obey the probability axioms), and even of degrees of holding outright beliefs, such as those discussed in Williamson [2000: 99], and of formulating a principle like (v) in terms of such things.

¹⁵ If one holds a threshold view of outright belief on which believing a proposition p outright just is assigning a high enough credence to p, then failing to give up one's belief in p will follow from failing to adjust one's credence in p. Then, (v) will entail (iv).

¹⁶ It follows that one can only ignore evidence against propositions that one believes, but since my concern here is with the idea that one is entitled to ignore evidence against propositions one knows, I won't worry about this.

¹⁷ This would be true even if believing h did require assigning it a credence of 1, since then in lowering one's credence in h one would also give up belief in h.

¹⁸ See Conee and Feldman [2004] for the position they call evidentialism.

¹⁹ See Conee [2004b], who also argues that accounting for knowledge defeat in this manner defuses the Dogmatism Puzzle.

²⁰ When I say that a reason R is defeated as a reason for believing p by other reasons the subject has, I mean that there are further undefeated reasons that have defeating force with respect to R as a reason to believe p, where this defeating force is so strong that the degree to which R supports p drops below the threshold required for knowledge (and justification). Note that this is not intended as a non-circular characterization of what it is for a reason to be defeated.

²¹ I do not want to fuss here about the difference between the evidence that a subject has and the reasons that a subject has, and sometimes talk just about evidence instead of evidence or reasons. However, I formulate this position by talking about reasons instead of evidence. This is because one might think that while items of evidence are themselves always certain, reasons need not be like this.

²² For instance, Pollock [1995] and Pollock and Cruz [1999]. For a critical discussion, see Lasonen-Aarnio [2010].

²³ For instance, one might take a subject's evidence to consist of externalistically individuated experiential states, such as the state of seeing a tree. If one takes evidence to consist of mental states, some analogue of the notion of entailment will need to be developed. For instance, one could suggest that being in mental state M entails h just in case, for some suitable notion of possibility, every possible case in which one is in M is a case in which h is true.

²⁴ Someone might object that not all propositions entailed by one's evidence are maximally supported by it. This would be to give up a probabilistic construal of the evidential support relation. Though I cannot argue for this here, a probabilistic relation strikes me as the most plausible, well-developed candidate for an objective relation, short of taking evidential support as a primitive relation. This conclusion is bolstered by the fact that in numerous simple examples (involving coin tosses, lotteries, etc.) our intuitions about evidential support obey a probabilistic structure. It is not surprising that most epistemologists who explicitly defend evidentialist views lean towards a probabilistic construal of the relation (see, e.g., Conee and Feldman [2004: 99–100, fn. 32], Conee [2004b: 264], and Feldman [2004: 117]).

²⁵ Consider a view on which knowing a proposition h requires not only that one's evidence entails h, but also that one knows that one's evidence entails h (thanks to a referee for this suggestion). One might hope, then, that the mere acquisition of new evidence could explain how it is that knowledge of h is defeated by explaining how it is that one's knowledge of the entailment is defeated. But, in effect, the same monotonicity problem arises, for if one knows that one's evidence entails h, then it must be entailed by one's evidence that one's evidence entails h, and merely adding new evidence cannot drop the degree to which one's evidence supports the proposition that one's evidence entails h. How, then, is the loss of knowledge that one's evidence entails h to be explained? The proposed view also raises a further issue, which is that to know a proposition h, a subject must know, for any number of iterations, that her evidence entails that her evidence entails … that her evidence entails h.

²⁶ What is much less straightforward is how to deal with defeat of necessarily true propositions, but I will set this problem aside.

²⁷ For instance, Conee [2004b], Pollock [1995], and Ginet [1980: 151].

²⁸ If one thinks that evidence consists of mental states, then assume that at t₁ the subject is in the mental state of seeing or seeming to see that draw 1 was red, etc.

²⁹ Even coherentists about justification will accept this, since whatever coherence among one's beliefs consists in, it doesn't consist merely in consistency.

³⁰ See, for instance, Williamson's [2000] anti-luminosity arguments.

³¹ A further problem with Veber's argument is that in defending his rejection of Entitlement, he assumes a view of what it is to disregard a piece of evidence E (as it bears on h) that I argued above is very problematic: disregarding E would be failing to come to believe or accept E in the first place. The resulting construal of Entitlement doesn't sound like something it is that bad to reject; it's not at all implausible to insist that sometimes one can justifiably come to believe or accept a proposition E even if one knows that it is misleading (as it bears on h).

³² Thanks to a referee for this suggestion.

³³ This article originates in the first chapter of my DPhil thesis, written in 2005. Many thanks to anonymous referees, to Sara Aronowitz, Daniel Morgan, Ram Neta, Tim Williamson, and Crispin Wright, and to an audience at a seminar at the University of St Andrews in 2008.