Epistemological Open Questions

Abstract

While there has been a great deal of recent interest in parallels between metaethics and metaepistemology, there has been little discussion of epistemological analogues of the open question argument (hereafter: OQA). This is somewhat surprising—the general trend in recent work is in the direction of emphasizing the continuity between metaethics and metaepistemology, and to treat metanormative questions as arising in parallel in these two normative domains. And while the OQA has been subjected to a wide variety of objections, it is still influential in metaethics. In this paper, I aim to show that an epistemological version of the OQA is just as promising as its moral cousin. That's not to say that I'll unqualifiedly endorse either argument. Rather, my aim is to show that there is just as strong a prima facie (really, secunda facie) case for an OQA in metaepistemology as there is in metaethics—I leave open whether the ultima facie case collapses.

Keywords:

• metaethics
• metaepistemology
• open question argument

Notes

¹ While the unreconstructed OQA is generally not regarded as persuasive, various descendants are still alive and kicking: especially, e.g. the moral twin-earth argument, due to Horgan and Timmons [1992]. Though see Dowell [forthcoming] for a recent critique.

² For example, OQA advocates who accept motivational internalism often cash out ‘openness’ in terms of a connection to motivation—this is historically how expressivists have interpreted Moore's argument. But those who reject motivational internalism while still offering variants of OQAs must understand openness differently. For example, Parfit [2011: ch. 27] argues against Schroeder's reductive naturalist view on broadly Moorean grounds: while he doesn't refer to ‘open questions’, he does accuse Schroeder of being unable to explain why certain questions are ‘substantive’, in a way that closely parallels Moorean complaints to the effect that reductionists cannot explain why certain questions are open. But Parfit is no motivational internalist, and he does not cash out the idea of ‘substantiveness’ in terms of a connection to motivation.

³ Though see Ridge [2007] and Heathwood [2009]. While Ridge doesn't put things explicitly in terms of OQAs, he does express sympathy for broadly Moorean, anti-reductionist, strategies for motivating epistemological expressivism [2007: 88–9]. Heathwood, whom I'll discuss below, argues explicitly against the parallel between moral and epistemic OQAs.

⁴ This view tends to be common ground among writers of a wide variety of metaethical persuasions. For some examples of recent treatments of metaepistemological issues in parallel to metaethical ones, see Cuneo [2007], Chrisman [2007], and Street [2009, 2010].

⁵ Darwall et al. [1992] understand openness in terms of intelligibility.

⁶ This is, for example, what Gibbard [2003] takes the upshot of OQA-style considerations to be.

⁷ This is how Horgan and Timmons [1992] develop their version of the OQA.

⁸ To say this would be to run into Williamson's [2007] objections.

⁹ The non-merely-epistemic notion of ‘counting against’ that is used here bears some similarity to the Wittgensteinian notion of a criterion, especially as developed by later commentators—see Lycan [1971] for a helpful survey.

¹⁰ See Gibbard [1990] for an expressivistic version of this strategy, and Wedgwood [2007] for a non-expressivistic one.

¹¹ As already mentioned, Parfit [2011] offers a version of an OQA, and while he doesn't defend any particular positive thesis about what it takes to have normative concepts, he explicitly argues against views of the sort mentioned in the previous note, on which normative concepts have constitutive connections to motivation.

¹² Thanks to an anonymous referee for raising this point.

¹³ See, e.g., Horwich [1998].

¹⁴ The claim that truth is a naturalistically specifiable goal might sound quite controversial. After all, the minimalist view about truth defended by Horwich ‘denies that truth and reference are complex or naturalistic properties’ [1998: 11]. The Quinean position in the text, however, is compatible with the minimalist view that truth is not a naturalistic property. It's enough for the Quinean argument to get going that questions about what is true are closed, once the natural facts are settled.

¹⁵ See Goldman [1994] for a discussion of the connections between various versions of reliabilism and naturalism.

¹⁶ See Berker [2013] for an interpretation and critique of reliabilism—and various other views in epistemology—as a version of epistemic consequentialism, with true belief as the goal.

¹⁷ Even Dummett didn't quite endorse an epistemic view of truth across the board. He allowed [2004] that propositions about the past might be true even though not supported by present and future total evidence.

¹⁸ For example, reliabilists identify justification, not with truth, but with the property of being produced by a belief-forming process that tends to lead to truth. See Goldman [1979].

¹⁹ To be sure, the strategy I'm suggesting is not uncontroversial. Writers in the ‘Knowledge First’ tradition, inspired by Williamson [2000], would argue that knowledge is the central object of study in epistemology and that, because knowledge is factive, questions about whether a subject knows that P are not logically independent from questions about whether P. Perhaps, for example, a belief is only fully justified when it is knowledge [Littlejohn 2012]. While the knowledge-first view does slightly complicate matters, I don't think that it blocks the basic sort of move I'm suggesting. As long as there is some normatively significant, positive, epistemic status a belief can have that is independent of its truth, some version of the strategy I've suggested can be employed. And knowledge-firsters will generally allow that such a status exists. For example, on Williamson's view, a belief's being epistemically probably true, given a body of evidence is logically independent of its truth. In that framework, our OQAs can be used to argue against reductive naturalistic analyses of epistemic probability.

²⁰ Exactly what it takes for one's credences to ‘match’ one's beliefs about chances is a matter of some controversy. A number of credence/chance linking principles have been proposed, the best-known of which is David Lewis's ‘principal principle’ [1980]. Exactly how we should express the right credence/chance link won't matter for my purposes.

²¹ ‘Tendency’ might suggest, not just ‘chance’, but a particular theory of chance—e.g. a propensity theory, of the sort typically associated with Popper [1957]. In fact, Goldman is explicitly neutral between understanding chances as propensities and understanding chances as frequencies [1979: 11].

²² To be clear, we can be sympathetic to reliabilism without holding that it provides a response to an epistemological OQA. We might think that reliabilism is correct as a first-order theory of justification, without holding that this has any special metaepistemological upshot, in much the same way that a moral expressivist or non-naturalist who accepts an OQA might still accept some particular first-order theory of ethics on which the moral actions are coextensive with some naturalistically specifiable set of actions.

²³ Goldman [1979] doesn't talk about naturalistic versus non-naturalistic claims, but he does distinguish ‘epistemic’ from ‘non-epistemic’ terms, and he is concerned to give an account of what it takes for a belief to be justified in entirely non-epistemic terms. Moreover, he takes it that ‘reliable’ is a non-epistemic term, as it is defined by way of non-epistemic terms like ‘tendency’.

²⁴ See, for example, Black [1998] and Hall [2004].

²⁵ See, for example, Forster and Sober [1994].

²⁶ While Lewis doesn't explicitly bring up curve-fitting, he does use a similar omniscient observer metaphor in explicating his account [1973: 74]: ‘Imagine that God has chosen to provide humanity with a Concise Encyclopedia of Unified Science, chosen according to His standards of truthfulness and our standards of simplicity and strength.’

²⁷ Strictly speaking, this will work only if the laws don't fix an end time for the universe. But I doubt this complication matters for present purposes.

²⁸ See Lewis [1980: 278–9] for an explanation of how his theory is, effectively, a special case of Jeffrey's.

²⁹ Here, ‘coherently fail to share’ should be read to mean that one might fail to share those attitudes without such failure counting against one's having a standard set of epistemic concepts.

³⁰ That's not to say that this is the only way of interpreting ‘best’ in ‘best system’ normatively. We might instead interpret the ‘best’ as, for example, ‘aesthetically best’. If this is how we read the best-systems account, then it will still be an open question whether one ought to conform one's credences to the chances according to the best system. Only when ‘best’ is given an epsitemological normative reading can we explain the apparent closedness of questions linking chances to rational credences.

³¹ See Greco [2014] for some relevant discussion. I argue that explaining the closedness of closely related questions is a serious challenge for many contemporary epistemological views, but that a certain sort of expressivism (of a sort that can also, I believe, take advantage of OQAs) is well-placed to provide an answer.

³² Thanks to Stephen Darwall, Tristram McPherson, Alex Silk, and Roger White, as well as two anonymous referees and the editors of the Australasian Journal of Philosophy for helpful comments and discussion.