1. Skepticism Untangled

Among the great doctrines of philosophy, none has seemed less attractive over the ages than skepticism. Indeed, for all the attention it has received, it is a view virtually undefended. Why this should be so is not hard to see. A skeptic, in the standard philosophical sense, is someone who denies that we have knowledge, either across the board, or at least with regard to a wide range of perfectly uncontroversial beliefs. This seems preposterous on its face, and matters only get worse in view of the consequence that might seem attendant on such a conclusion: that if we do not know, then we ought not to believe. Abandoning a large proportion of our beliefs is an outcome that no one can reasonably be expected to embrace. Accordingly, a 2009 PhilPapers survey (philpapers.org/surveys) found that less than 5 percent of professional philosophers identified themselves as embracing or even leaning toward skepticism regarding the external world. More than 80 percent embraced antiskepticism: a greater degree of consensus than was found on any of the other 29 questions asked.

Things look different, however, if we conceptualize this territory in other ways. Reflection on the history of skepticism shows that philosophers have often conjoined as a single doctrine various theses that are best kept apart. Some of these theses are incredible—literally almost impossible to accept—whereas others seem quite plausible and even verge on the platitudinous. Mixing them together, one arrives at a view—skepticism—that is as a whole indefensible. My aim is to pull these different elements apart and to focus on one particular strand of skepticism that deserves sustained and respectful attention.

The strand I have in mind I will refer to as epistemic defeatism. Roughly, in its most global form, it is the view that, in the final analysis, we have no good evidence for the truth of any proposition. Put a bit more carefully, we can think of ‘good’ evidence as evidence that is undefeated, where evidence can be defeated by being either rebutted by further evidence for denying the proposition or undercut by further evidence that undermines the force of the original evidence. Epistemic defeatism holds that ‘ultimately’ there is no good evidence in the sense that, for any subset of evidence that supports a proposition, additional evidence is always available that, in the end, wholly defeats the original evidence, entirely canceling its evidential weight. Epistemic defeatism need not deny that, in actual practice, given the many assumptions we make about the world, we have a great deal of evidence. But the view insists that such evidence is not ultimately good, because it is always conditional on taking for granted certain things, or ignoring other things, and that these assumptions, tacit or explicit, cannot themselves ultimately be supported by good evidence.

I will not attempt to argue for epistemic defeatism in this paper. It is, under one description or another, as well-known as any idea in philosophy, and many of the arguments that might be wielded in its support are familiar enough that they require no detailed discussion here. I also will not try to defend defeatism against the most obvious objections, such as its worrisome appearance of being self-refuting. In the end, I am myself unsure whether I find the view compelling, though I do think it has, on its face, considerable plausibility, even when construed globally. (Since I will not be attempting to defend the thesis, I also will not bother to distinguish its more or less global versions—external world, rational intuition, morality, and so forth—and will instead focus on the most general version of the thesis, globally extended to all domains.) Despite its familiarity, epistemic defeatism lives in the philosophical shadows. Occluded by our preoccupation with the concept of knowledge, it is a view that goes by no name and has no standard formulation. What I will first attempt, then, by working through various historical texts, is to untangle epistemic defeatism from neighboring views and in particular to establish its independence from questions about knowledge. Having thus established the view's autonomy, I will turn in the second half of the paper to considering the options for self-consciously accepting defeat. One may despair, or one may have faith. But I will ultimately propose that the most attractive option—the option that preserves the most of our epistemic integrity—is to have hope.

2. Two Historical Paradigms

Although overshadowed in recent decades by debates over knowledge, in previous eras epistemic defeatism has received several very prominent and explicit defenses. The first is that of the ancient skeptics, best known today under the banner of Pyrrhonism, whose cardinal doctrine is that ‘to every account an equal account is opposed’ (Sextus Empiricus, Outlines I.6.12). Admirably careful to avoid any positive assertions that go beyond the appearances, the Pyrrhonians offer this only as a report about how things seem to them. As Sextus Empiricus more carefully puts it, ‘to every account I have scrutinized … there appears to me to be opposed another account, … equal to it in convincingness or lack of convincingness’ (Outlines I.27.203). Of course, as the Pyrrhonians knew perfectly well, this is not how things appear to the vast majority of us, entrenched as we are in our dogmatic worldviews. But the idea for which they tirelessly argue is that if we step back from our particular worldviews and consider without prejudice the totality of the evidence, all the way down to the ground, then in fact nothing is more likely to be the case than anything else. The repertoire of arguments employed by the Pyrrhonians—their ‘modes’—is vast, and includes the now familiar arguments from illusion, perceptual variation, disagreement, and cultural relativity. Perhaps their most powerful general argument, however, forces its opponent into one of three bad options: an infinite regress of premises, circularity, or groundless assumptions (on these so-called Five Modes of Agrippa, see Barnes [Reference Barnes1990] and Fogelin [Reference Fogelin1994]). Since these are alleged to be the only options, and since each is said to undermine the cogency of any reasoning, the result is epistemic defeat. On the Pyrrhonian formulation, no reasoning appears any more likely than its opposite.

Through the Middle Ages it is rare to find defenses of any of the various strands of skepticism. Famously, however, such ideas become more influential in the sixteenth and seventeenth centuries (see e.g., Perler [Reference Perler and Pasnau2010] and Popkin [Reference Popkin2003]). A Latin translation of Sextus's Outlines of Skepticism appears in print in 1562, and by the time of Descartes, nearly a century later, the full repertoire of skeptical arguments has become tediously familiar. These trends culminate, a century later still, in the work of David Hume, the second great proponent of epistemic defeatism. In Hume's words, ‘’tis impossible upon any system to defend either our understanding or senses’ (Treatise I.4.2: 218). For Hume, the principal argument is straightforward: to defend either understanding or the senses, we would need some way to rise up and evaluate the reliability of those faculties. But since any such evaluation presupposes the very faculties that are being evaluated, the project is hopelessly circular. As a result, we are left with ‘not the lowest degree of evidence in any proposition, either in philosophy or common life’ (Treatise I.4.7: 267–68).

3. Weak versus Strong Skepticism

Before considering the prospects for knowledge in the face of epistemic defeat, it will be helpful to look at how arguments for skepticism can run with or without defeatism. Call strong skepticism the denial of knowledge grounded in epistemic defeatism. Weak skepticism, in contrast, is the view that allows our beliefs to have some positive evidential support but denies that this support is sufficient for knowledge (for this distinction see, e.g., Huemer Reference Huemer2001: 20–21). Weak skepticism, we can say, is based on a weak version of epistemic defeatism according to which all of our beliefs are vulnerable to some measure of defeating evidence, but yet may still manage, ultimately, to retain positive evidential support. Strong skepticism is based on the full-fledged version of epistemic defeatism according to which our beliefs are so thoroughly defeated that they ultimately lack any positive evidential support.

The best-known proponent of weak skepticism is Descartes. In the midst of the First Meditation, he interrupts his recitation of skeptical arguments to remind the reader that the everyday beliefs he aims to subvert are ‘highly probable’ and ‘much more reasonable to believe than to deny’ (VII:22). In effect, Descartes wants to put one foot into the Epistemology Room, where all beliefs get heightened scrutiny, even while he keeps his other foot in the ordinary world. Out in the world, our regular beliefs about the external world are, of course, ‘much more reasonable’ than not. But the ambition of the Meditations is to elevate such ordinary beliefs to the exalted level of scientia, thereby achieving ideal knowledge regarding God, the soul, and physical bodies (for a more detailed discussion of Descartes's aspirations to scientia, see Pasnau [Reference Pasnau2014]). His strategy to achieve that goal is to lock himself and his readers into the Epistemology Room for the duration of six meditations. The skeptical scenarios that initially motivate the project do not yield epistemic defeatism but just the slightest basis for doubt—enough to deprive us of certainty.

Although weak skepticism has had its modern defenders,Footnote ¹ most philosophers today seem to find the position tedious. Of course there is nothing, or almost nothing, that we should adhere to with complete certainty. Does this mean we know nothing or almost nothing? Really, if that is all that is at stake in the quarrel over skepticism, then who cares? My suspicion is that weak skepticism owes its prominence in philosophy today largely to a historical misinterpretation or, more precisely, a mistranslation. We see a long line of philosophers insisting, quite reasonably, that the ideal goal of inquiry is certainty, using words like ‘epistēmē’ and ‘scientia’ and even ‘knowledge’ as labels for that ideal. Reading these texts as if they were written in modern English, we thus find them saying that ‘knowledge requires certainty’, and so we suppose that if skepticism is to be kept at bay, certainty must be achieved. If anyone wonders why the standards for knowledge must be held so high, the response is just that this is what the word means. But if the word really has to be given over to some obsessive-compulsive quest for certainty, then the appropriate next step is to find another, more tractable word and move on. The more interesting form of skepticism is grounded in epistemic defeatism. For if, in the final analysis, we have no good evidence for the truth of any proposition, then it may seem inevitable that we lack knowledge.

4. Knowledge

The inference from epistemic defeatism to the skeptical denial of knowledge has been almost universally treated as a reductio: we obviously do have knowledge, and so defeatism is absurd. Once dismissed through this oblique strategy, epistemic defeatism recedes into the shadows, not because it has been directly refuted, but just because it seems to be of interest only for as long as the skeptical denial of knowledge is of interest.

The correct conclusion, however, is not that epistemic defeatism is untenable, but rather that we can perfectly well have knowledge even if epistemic defeatism is true. Indeed, as anyone with a passing familiarity with recent epistemology will know, there must be a hundred ways of getting this result. One might be a reliabilist about knowledge, supposing that a true belief formed through a reliable mechanism counts as knowledge, even if we cannot show the mechanism to be reliable (e.g., Goldman Reference Goldman and Pappas1979). One can suppose that the true beliefs we form have the default status of being known until proven otherwise (e.g., Harman Reference Harman and Luper2003). One can take questions of knowledge to be answerable in terms of language games that give such questions meaning (Wittgenstein Reference Wittgenstein, Anscombe, von Wright, Paul and Anscombe1969) or in terms of changeable contexts that ordinarily allow for fairly lax standards (e.g., Cohen Reference Cohen1986). One can count as known the consequences of selected unargued premises (perhaps Pryor Reference Pryor2000). One can appeal to coherence (e.g., Lehrer Reference Lehrer1990). And on and on.

By no means do I wish to take for granted that one or another of these defeat-friendly theories is ultimately successful in rescuing knowledge from the jaws of epistemic defeat. Clearly, defeatism poses a serious prima facie threat to our possession of knowledge, and any attempt to grapple with it must have something to say about the kinds of defeat that threaten knowledge and the kinds that do not. Here all I mean to assert is that we should not assume that a satisfactory account of knowledge inoculates us against a verdict of epistemic defeat. For even if we embrace one or another of these relaxed conceptions of knowledge, independent questions remain about whether we ultimately have good evidence for our beliefs or whether we inevitably must beg the question, argue in circles, or commit some similar epistemic sin.

The earlier history of epistemology bears out the impression that these different strands of skepticism need to be kept apart. To be sure, the great historical proponents of defeatism connect it with skepticism in our sense of the word. Both the Pyrrhonians and Hume think defeatism rationally demands the suspension of belief, which would seem to leave no room for anything very close to our concept of knowledge. Yet it is striking, all the same, how disconnected their discussions are from any such concept. The Pyrrhonians, insofar as their work survives, do not frame their account in terms that are at all close to our terminology of knowledge. Indeed, the standard Annas–Barnes English translation of Sextus's Outlines uses the word ‘knowledge’ and its cognates only a handful of times and never in a technical sense. Hume does talk about knowledge, but he uses the word in its traditional sense for an ideally certain cognitive state, which he allows only when we compare ideas (Treatise I.3.1: 69–70). When he argues for epistemic defeatism, he is aiming to show not that we lack ‘knowledge’—an all-too-easy target, by his lights—but that our beliefs lack even probability.

Epistemology today, in contrast, fixated as it is on the possession of knowledge, has tended to ignore the thesis of epistemic defeatism, showing interest in it only as a premise for skepticism. It is as if, once a way is found to secure knowledge, defeatism can be safely forgotten, even if it has not been refuted. As a result, we ignore one of the most fundamental questions we might ask about our place in the world. After all, regardless of how we use the word ‘knowledge’, what could possibly be of greater philosophical significance than the thesis that, in the final analysis, we have no good evidence for the truth of any proposition?

5. Platitudinous or Incoherent

This last question is not merely rhetorical. Once epistemic defeatism has been distinguished from questions about knowledge, there is room to wonder whether it actually amounts to a substantive thesis at all. From one side, the worry might arise that epistemic defeatism is nothing more than a platitude, something in the vicinity of saying that all arguments have to start somewhere. And while perhaps there were philosophers of old who supposed that the starting points of arguments could themselves somehow be made to stand on their own as wholly self-evident, surely today it is hardly newsworthy to decry the demise of foundationalism. Indeed, this was a familiar point even in antiquity. As Aristotle put it in his remarks against the skeptics, ‘their mistake is that … they seek a reason for that for which no reason can be given; for the starting-point of demonstration is not demonstration’ (Metaphysics IV 6, 1011a12–13).

But epistemic defeatism says more than that foundationalism is an impossible dream. It denies across the board the possibility of any successful strategy for defending our beliefs in terms of objective reasons that are not ultimately based on one or another objectionable epistemic practice. Hence the theory is couched in quite general terms as our inability, through any means, to acquire ultimately good evidence for our beliefs.

Yet, as soon as one attempts to describe epistemic defeatism in broad epistemic terms such as ‘evidence’, worries begin to arise over whether there is a cogent thesis to be had here at all. For instance, anyone who would defend our possession of knowledge in terms of the reliability of our cognitive faculties is likely to think that reliable faculties are, all by themselves, a source of good evidence.Footnote ² When evidence and knowledge are conceived in this way, knowledge precludes epistemic defeat. Moreover, although there are other ways in which one might formulate defeatism—for instance, in terms of justification, objectivity, or reasons—it seems likely that these alternatives will be prone to similar sorts of complaints. In general, the problem is that for every term of epistemic art that one might want to rely on in formulating the thesis of epistemic defeatism, there exists an interpretation of that term that has been carefully framed to track the possession of knowledge. The lesson to be learned from this, however, is not that epistemic defeatism, when paired with the avowal of knowledge, becomes incoherent. Instead, the conclusion to be drawn is just how conceptually entangled the various strands of skepticism have become.

The high road to untangling these strands would be to build from the ground up an epistemic framework suitable for making the distinctions I am after, defining along the way the notions of evidence and knowledge. This, alas, is not that paper. Efforts along these lines certainly have been made, most prominently by distinguishing between internalism and externalism in epistemology. Epistemic defeatism could thus be formulated in terms of an internalist theory of evidence (e.g., we have to be able to show that our faculties are reliable) while allowing for knowledge in externalist terms (e.g., those faculties just have in fact to be reliable). But rather than commit myself to any such particular path, which would be bound to be contentious, I will continue to follow the low road of appealing to historical precedents, an approach that I hope has the advantage of displaying the issues in their ageless generality.

When one does attend to the historical record, it is very hard to believe there is no coherent and substantive thesis in the vicinity of epistemic defeatism. The Pyrrhonian challenge described above—that arguments will be either circular, infinite, or arbitrary—is one that philosophers over the centuries have long sought to meet. Aristotle himself, at the end of the Posterior Analytics, attempts an account of what grounds the principles of demonstration. If the ultimate groundlessness of all belief were a truism or incoherent, then it would be very strange that the problem has preoccupied philosophers so deeply and for so long. This suggests that, even if we accept one or another relaxed modern standard for having knowledge, there remains something else we might want to have, something that the epistemic defeatist maintains we cannot have.

6. Rational Expectation

Another way of conceiving of epistemic defeatism, one that will be helpful later, is to draw on the notion of rational expectation. Here we might go back to one of the first great treatments of probabilistic reasoning, from around the end of the seventeenth century, Jacob Bernoulli's Art of Conjecturing. Bernoulli's interests are mainly mathematical: he seeks precise quantitative measurements of what expectation an agent should have in various sorts of situations, particularly situations involving games of chance. To take one of his simplest examples, if I am to choose either what is in your left hand or your right hand, and I know there are seven coins in one hand and three in the other (but I do not know which amount is where), then I can expect (7+3)/2, and so my expectation should be five coins. But what is an expectation? According to Bernoulli (Reference Bernoulli and Sylla2006: 134):

Bernoulli is describing an epistemic ideal that we would now characterize in terms of rationality. Although you might hope for seven coins, and although you might fear receiving only three, what you can rationally expect is five. In these terms, global epistemic defeatism is the view that we are never in a position to weigh our hopes and fears and thereby arrive at some rational expectation in between. We may fear systematic deception, and we may hope for reliable veridicality, but since there are no valid grounds to weigh the probability of any outcome, there is no rational basis to adopt any credence between 0 and 1. There are ultimately no grounds, so far as we can tell, for any expectation.

It is tempting to think of epistemic defeatism as prescribing a credence of .5 in every proposition—a state of perfect indecision. But this leads to trouble. Even an epistemic defeatist should be worried about a view on which your credence in each of the following is .5: the ball is red; the ball is scarlet; the ball is green; the ball has no color; there is no ball. As Bas van Fraassen (Reference van Fraassen1989) establishes, agnosticism is better represented in formal epistemology by a set of probability functions, so that an undecided agent's degree of belief is consistent with a range of probabilities. For the epistemic defeatist, the interval would presumably run all the way from 0 to 1, meaning that, in effect, the defeatist whose credences are determined purely by the evidence will have no definite credences.

Since Bernoulli's time, formal epistemology has been developed to a high pitch of sophistication, relying particularly on the fundamental work of Thomas Bayes. But Bayesianism presupposes a set of prior credences on which it can build. The essence of epistemic defeatism is to press the question of what we have reason to believe if we disallow all such assumptions and biases. From such a starting point, the defeatist contends, we would have no basis for taking our expectations in one direction or another. The familiar heuristic here is that of a wager. In its global form, epistemic defeatism asserts that, ultimately, no advantage is to be expected from wagering on anything. In short: no wager is rational. No platitude here—this is truly calamitous. So much so, indeed, that it looks patently absurd. Who, after all, would turn down a chance to put money on the sun's rising in the morning? But the absurdity arises because we imagine making wagers in a context where both parties accept certain presuppositions—about what would count as confirming a sunrise, about what ‘sunrise’ means, about how wagers work, and so on. If I offer you this bet, but then carefully explain that no such presuppositions will be allowed in determining the winner, you may well decide that the bet has no value.

7. Antirealism

Another way to articulate epistemic defeatism is to compare it with various forms of antirealism. Consider, first, the analogy to ethics. The moral realist thinks that the rightness of an action has some kind of objective ground, something that makes an action morally good, independent of contingent facts about what human beings happen to care about. For the moral antirealist, in contrast, there is no such objective ground. If we can aptly speak of moral rightness at all, it is a function only of what we in fact happen to value. Epistemic defeatism makes analogous claims. Deniers of epistemic defeat hold that there are ultimate, objective evidential grounds that make some beliefs more rational than others. According to the epistemic defeatist, in contrast, if one can aptly speak of beliefs as being rational at all, they must ultimately take their rationality from subjective facts about what believers are inclined to think. No objective evidential grounds are available. One might even use the label ‘epistemic antirealism’ for what I am now calling epistemic defeatism. But ‘antirealism’ in the epistemic domain is best reserved for views that argue for expressivism or some other form of antirealism with regard to specifically epistemic norms, such as We ought to believe in accord with the evidence (see Field Reference Field2009). Still, there is a close connection between moral antirealism and epistemic defeatism. For just as, in the moral domain, Hume denied there could be any argument running from is to ought, so the epistemic defeatist despairs of our reasoning from seems to is.

There are also interesting connections with metaphysical antirealism. Berkeley, for instance, opts for metaphysical antirealism precisely in order to forestall the risk of epistemic defeat. The realist who embraces the existence of things beyond our ideas—that is, a reality absolutely independent of our minds—is ‘thrown into the most hopeless and abandoned scepticism’ (Three Dialogues, dial. 3, II:246). The only way to escape epistemic defeat, for Berkeley, is to be a metaphysical antirealist, at least with regard to the material realm, thereby rendering the domain of what is coextensive with the domain of what seems.

Another instructive case is Kant, who in effect replaces Berkeley's metaphysical antirealism with a distinction between the empirical and the transcendental, allowing epistemic success in the first case but insisting on defeat at the transcendental level. Depending on one's sympathies, his distinction between the phenomenal and the noumenal amounts either to a brilliant escape from the seems–is gap, or merely to an alternative description of it. The labyrinths of the first Critique yield a systematic account of our seemings—of what is empirically real—in consolation for our inability to say anything about what is transcendentally real. Kant thus offers a way forward in philosophy that fully honors epistemic defeatism. Even without making claims about the mind-independent world, there remains a great deal for the philosopher to say.

8. Logical Defeatism

It is a commonplace, over the centuries, to find our limited epistemic prospects contrasted with the unlimited perfection of God. True certainty, it is regularly said, is possible only in the divine case, leaving us in possession only of a lesser sort of human certainty. It is natural, then, to think of epistemic defeatism as specifically endemic to human beings—as Quine memorably quipped, ‘the Humean predicament is the human predicament’ (Quine Reference Quine1969: 72). Other rational beings, one might imagine, are not so limited in their powers. An all-powerful deity, presumably, would face no such limits.

Yet there is reason to suspect that epistemic defeatism, if true, is a logical truth that applies to all cognitive beings. For consider even the case of God. Presumably it seems to God that he is the omnipotent, eternal, creator of the universe. But might he be mistaken? Might he be dreaming? Might some evil demon be tricking him into falsely thinking he is God? No, not if he is God, but that is precisely what is in doubt. To suppose that God might have such doubts may seem absurd, but the suggestion ought to be taken seriously. After all, there is nothing about epistemic defeatism that turns on distinctive features of the human epistemic predicament. The structure of the arguments depends on wholly logical points about evidence and confirmation. Hence the defeatist's arguments, if they have force at all, would seem to have just as much force against God as they do against human beings. That is why the proponent of defeatism feels so confident that—no matter what line of reasoning might be advanced—we can never ultimately arrive at any noncircular basis for bridging the gap between seems and is. The impossibility holds as a matter of logic, rather than as a matter of our contingent epistemic condition (compare Nozick Reference Nozick1997).

9. Pyrrhonian Suspension

I have not argued for epistemic defeatism, only sought to make its character clear. Let me now turn to the second half of my agenda and consider epistemic defeatism's consequences for our epistemic practices. Historically, the dominant thought has been that it leads to the most radical of results: that we should suspend belief across every domain where epistemic defeatism holds. This is how ancient Pyrrhonism is generally understood. According to Sextus, the Pyrrhonists ‘say what is apparent to themselves and report their own feelings without holding opinions, affirming nothing about external objects’ (Outlines I.vii.15). There is considerable scholarly debate over just how to understand such remarks and over whether the Pyrrhonians truly propose to suspend all belief (see Frede and Burnyeat Reference Frede and Burnyeat1997). Without taking a side on the details of that exegetical dispute, I will, for the sake of convenience, use the label ‘Pyrrhonian’ to mark out this first, most extreme way of reacting to epistemic defeat: as requiring the suspension of belief.

Following the usual understanding of Pyrrhonism, Hume takes epistemic defeat to entail that we should suspend belief. But Hume's view is different in two crucial ways. First, the Pyrrhonians treated their view not as an intellectual calamity, but rather as the best way to a happy life. Sextus, in fact, thinks the whole aim of Pyrrhonism is the achievement of tranquility. For Hume, in contrast, Pyrrhonism is rather ‘the most deplorable condition imaginable’ (Treatise I.4.7: 269). The only consolation Hume takes is that we are by nature incapable of adhering to the Pyrrhonian program. This is the second crucial difference. Although we may be able to turn our study into an Epistemology Room for a few hours, Hume thinks we cannot stay there for long, and that when we emerge, we inevitably return to our familiar beliefs. Accordingly, he refers to ancient skepticism as a ‘fantastic sect’ (Treatise I.4.1: 183) and calls its proponent an ‘absurd creature’ (Enquiry conc. Human Understanding 12.1: 149) who need not be refuted because the skeptic's claims are impossible for us to accept.

Philosophers largely continue to take Hume's view that there is something absurd about skepticism. But the absurdity resides just where Hume says it does, in the injunction to suspend belief, rather than in the epistemic defeatism that lies behind the view. So we should consider whether there is some way to embrace defeatism without following the Pyrrhonians all the way to their supposedly tranquil end.

10. Evidentialism

Hume himself is of limited help in thinking about how to escape epistemic defeat. To be sure, he has a great deal to say about the sources and structure of the beliefs we do hold. Just as his moral antirealism accompanies well-developed ideas about our moral practices, so his epistemic defeatism stands alongside an extended account of our cognitive psychology. Even so, Hume's attitude toward epistemic defeat is one of despair. Rather than offer a coherent defense of our doxastic practices in light of epistemic defeatism, he simply registers the fact that we continue to form beliefs, and he accordingly turns his attention to an account of their nature. As he says, with regard to his defeatist conclusions, ‘I know not what ought to be done in the present case. I can only observe what is commonly done; which is, that this difficulty is seldom or never thought of; and even where it has once been present to the mind, is quickly forgot’ (Treatise I.4.7: 268). Little has changed in this regard since Hume's day. It is rather like our attitude toward death: we know what is coming, but most of us are by nature mercifully unable to dwell on it. (Not everyone takes Hume's remarks on skepticism to clash with his positive cognitive theory. For a recent attempt at reconciliation, see Schmitt Reference Schmitt2014.)

What seems to lead Hume to despair in epistemology, but not in ethics, is his tacit acceptance of evidentialism, by which I mean the thesis that we should form a belief only when we possess sufficient evidence. Since epistemic defeatism contends that, ultimately, we never possess sufficient evidence, the addition of evidentialism yields the Pyrrhonian result that we should suspend belief. Given the prima facie plausibility of evidentialism, Hume's despair looks reasonable, and one can see why epistemic defeatism has generally been conflated with full-blown skepticism. Yet, at the same time there is something dubious about the conjunction of evidentialism and defeatism. Evidentialism looks plausible because it describes an ideal that has tremendous appeal. No one could deny that it would be a very good thing if all our beliefs could be grounded in sufficient evidence. So, if this is something we are capable of doing, it is natural to suppose that we ought to do it. But what if, as epistemic defeatism holds, this is not something we are capable of? What happens when we squarely acknowledge the reality of epistemic defeat? Then we face a much harder choice: either abandon belief or abandon evidentialism and tolerate beliefs formed on some less demanding grounds. It is as if your mother were to recite to you that eminently sound piece of advice, ‘If you cannot say anything nice, do not say anything at all’, and you were subsequently to conclude, on reflection, that you had nothing nice to say. Should you keep silent, indefinitely? Or should you reject your mother's advice?

The choice is not obvious, neither in epistemology nor at the kitchen table. Here we can see the force of Pyrrhonism, but at the same time we can see its absurdity: that one would take the normative principle of evidentialism so seriously as to bind oneself to a lifetime of suspending belief. At this point the Pyrrhonians have several further things to say. First, they think that they can go on with their ordinary lives by following the appearances without assenting to them. Second, as already noted, they have an impetus to attempt the suspension of belief that goes beyond a compulsive adherence to evidentialism: for they take themselves to be pursuing a strategy that will bring them tranquility. But if we do not think that suspending belief is likely to yield happiness or that a life without belief is livable, then we ought to take seriously the strategy of replacing evidentialism with some less demanding rule for belief formation. Yet what could that be?

Hume's despair is, in effect, one kind of solution. He offers a form of philosophical quietism, according to which we just carry on as nature intended, without attempting a theoretical resolution. But our challenge is to do better. It might seem that the various defeat-friendly theories of knowledge canvassed in §4 above offer some help here, inasmuch as each points toward one or another strategy for resisting evidentialism. But these theories on their own do not answer the question before us. Sometimes they assume that we have at least a measure of good evidence for our beliefs, and so such theories touch only the weak version of epistemic defeatism that gives rise to weak skepticism (as we saw in §3 above). Other theories concentrate on naïve subjects who have never considered the sort of defeating considerations that the skeptic puts forward, or they focus on conversational contexts where those skeptical concerns are not relevant. The challenge before us now is more difficult: how can someone explicitly embrace epistemic defeatism in its strong form and yet continue to form beliefs in the way we all ordinarily do? Since knowledge presumably requires belief, none of the defeat-friendly theories of knowledge will be of any help here—not until we have some self-consciously defensible response to defeatism other than Pyrrhonian belief-suspension.

11. Divine Inspiration

The questions I am asking have received some attention in religious contexts, a domain where Pyrrhonian suspension of belief is not an option, but where the sufficiency of our evidence is regularly doubted. Augustine, for instance, begins his Soliloquies with a prayer that he may find God in one way or another:

Ideally, Augustine would like to achieve understanding of God and our relationship to him, but he does not assume that this will be possible, and so he asks for whatever works, whether that be knowledge, faith, or virtue. His problem is essentially epistemic, but he is open to a range of solutions beyond simply the accumulation of evidence in the pursuit of certainty.

It is easy to multiply examples. Al-Ghazālī describes having suffered through a skeptical crisis that came to an end only when God ‘cured’ him. ‘This did not come about by composing a proof or by arranging some words (), but rather by a light that God Almighty cast into my breast, this light being the key to the greater part of knowledge ()’ (Rescuer from Error n. 86, tr. Khalidi Reference Khalidi2005: 63, modified). Another example is Pierre Bayle, someone whose broader views—skeptical and cynical—could scarcely be more different from Ghazālī's or Augustine’s. After describing with some enthusiasm the doctrines of Pyrrhonism, Bayle remarks that ‘the natural conclusion of this ought to be to renounce this guide [i.e., reason] and to implore the cause of all things to give us a better one’ (Reference Burleigh1991: 206).

Yet it is not clear that even divine assistance can help. To be sure, God might, in any number of ways, instill within us confidence about how things are. But this is not the help we need: as Hume stresses, nature itself takes care of sustaining our confidence. The sort of divine help that would be useful is a strategy for defeating the ultimate defeaters of epistemic defeatism. But if defeatism holds as a matter of logic (see §8 above), then not even God could give us this, because God himself is not able to transcend the seems–is gap. This is not to say that God could not teach us anything. The point is just that God is in no better position than we are to deal with the foundational problem posed by epistemic defeat. As far as that goes, appealing to God is no better than appealing to the Wykeham Professor of Logic.

12. Faith

Even if God cannot provide the ultimate cure for epistemic defeat, Augustine's invocation of the theological virtues points toward the two most promising strategies for reconciling defeatism with a nonskeptical, nonidealistic worldview. First, let us consider faith, and then hope (saving charity, as usual, for another day).

To hold a belief on faith, as I use that term, is to hold it firmly and attach a high credence to it even though one does not suppose that the evidence warrants such confidence. Believing on faith, so understood, directly clashes with evidentialism, and it strikes many as an extremely disreputable option. Some have argued that this kind of fideism is incoherent in the sense that no agent could self-consciously maintain one credence while recognizing that the evidence warrants a different credence (e.g., Adler Reference Adler2002). Others would simply point to the badness of believing things that are insufficiently supported by evidence (e.g., Clifford Reference Clifford1999). But the epistemic defeatist will reply that such charges need to be evaluated in light of the other options we face. To be sure, ours would be a better world if we never needed to take anything on faith. But, says the fideistic epistemic defeatist, our world is not so ideal. We must work with what we have. If the remaining options are some kind of metaphysical antirealism or else Pyrrhonian belief-suspension, then fideism may be the least bad alternative available.

It seems natural to develop fideism in geometric terms. That is, we would consider which axioms—not self-evident axioms, as Euclid aspired to, but unargued axioms, taken on faith—might yield the most comprehensive system of beliefs. In God's case, this is easy: the only axiom God needs is that he is God. From this, the rest follows. In the human case, it is less clear how to proceed. Some have taken as axiomatic the analogous principle that we are the creations of a good and all-powerful God. Others have proposed to cross the gap between seems and is directly by embracing a principle according to which a thing's seeming to be so yields prima facie good evidence for taking it to be so. Once we let faith into the game, interesting questions arise about how much one ought to take on faith. The geometric model suggests that parsimony will weigh quite heavily here, and that the ideal theory will take for granted as little as possible. But one might think that it scarcely matters whether we believe just one thing on faith or dozens of things. After all, if even one faith-based proposition is foundational for the whole system, then there is an obvious sense in which everything is based on faith. And if this is how things are, then perhaps we should take seriously the way in which faith seems to play a role in our everyday epistemic practices: our faith in friends and family, even faith in oneself. Clearly, however, this is a dangerous path, for once the constraints are loosened on what may be embraced through faith, we risk epistemic chaos. How any constraints might be maintained is a central challenge for the fideistic defeatist.

13. Hope

A less obvious path toward epistemic defeatism—but one I regard as preferable—relies not on faith but on hope. To hold a proposition on hope, as I use the term, involves believing that proposition without having a high credence in its truth. To make sense of this takes a bit of work. First, we must insist on something I have already taken for granted at various points, namely, a distinction between credences and absolute beliefs. A credence is an attitude regarding the chance of a proposition's obtaining. A belief is something different: it might be understood in terms of behavioral and cognitive dispositions, in particular, a disposition to assert the proposition in question (see e.g., Frankish Reference Frankish, Huber and Schmidt-Petri2009). Next, we must treat these two kinds of doxastic states as capable of independent variation. So, today, I might have a credence of .7 in one proposition and believe that proposition. Or I might have a credence of .8 in another proposition and yet not believe it. Tomorrow, credences unchanged, I might cease believing the first proposition and begin believing the second. With this distinction in mind, consider an agent who endorses epistemic defeatism and who refuses to allow her credences to be swayed by faith. It seems that she should have no definite credence in any proposition. Does this require Pyrrhonian suspension of belief? No. If beliefs need not track credences, she still might believe.

Why would such an agent form beliefs? She might do so out of hope. To understand this takes some more work. Recall Bernoulli's earlier characterization of expectation as lying halfway between hope and fear (§6 above). His formulation self-consciously builds on earlier scholastic accounts, which conceive of the presence and absence of fear as a defining feature of our various doxastic states. According to Albert the Great, for instance, ‘opinion, when it is inclined to a thing by [merely] probable reasons, fears the contradictory on account of the weakness of its reasons’ (Ethica VII.1.2.5 [Opera VII:467]). Scholastic discussions standardly distinguish two ways of escaping from this state of doxastic fear. One way is to acquire sufficient reasons, in which case one can count as knowing a proposition. The other way is to embrace the proposition on faith. But, once we distinguish between credence and belief, a third possibility opens up: that without elevating one's credence through faith, one might simply stop fearing that one is wrong. Instead of fearing and hoping in equal measure, one places one's hopes entirely in being right, and so believes.

It may seem strange to imagine hope and credence parting ways like this, but in fact this can readily happen. The critical thing to notice is that hope and fear are affective rather than cognitive states. As such, they not only can but regularly do run in different directions from our cognitive judgments. One may have an extremely high credence that a thing is so and yet fear that it is not. Or one may have a fairly low credence in a proposition and yet regard it with sunny optimism. As an illustration of the role such affective attitudes can play in shaping belief, consider one of the very few medieval expressions of skepticism, that of Francesco Petrarch:

Giving greater weight to the risk of error, Petrarch takes the side of fear rather than hope. This leads him to a skeptical suspension of belief—nearly a global skepticism, but for his prudent exemption of religious belief. Yet, we can imagine that, without any change on the cognitive side in his credences about the plausibility of the skeptical arguments, he might instead have taken up an attitude of hope. This would mean setting aside his worries about being ‘entangled in error’, and optimistically dwelling on the prospects of being right. To become optimistic in this way does not require supposing it more likely that one is right: the initial change required is affective rather than cognitive. Where optimism is diluted by fear, according to the medieval analysis, there can at most be uncertain opinion. Hence Petrarch is a skeptic. But where there is hope, fear is assuaged, and hence there can be confident belief. Here, then, is a way to combat Pyrrhonian belief-suspension indirectly, not by rejecting epistemic defeatism but simply by ceasing to worry about it.

At this point, a fuller account would need to explain in some detail what it is to have a belief and then give a distinct explanation of what it is to have a credence. One would then be able to see exactly how an agent could have belief by virtue of sunny optimism while yet refraining from adopting any particular credence. This is a vulnerable point in my story, because one might naturally suppose that optimism is the very sign of an agent with high credences. For instance, credence is often understood in terms of betting behavior, and one would expect such behavior to be influenced by optimism. My sketch of an account exploits the connection between belief and affect, arguing that optimism as an affective state can give rise to belief. That suggests there can be another part of the agent's psychology where the credences remain unchanged. Here is where a fuller account would need to say more.

As an example, think of a professional gambler whose affective states swing wildly from night to night, with the result that sometimes he believes he will win and other times he believes he will lose. Even so, professional that he is, he deploys the same strategy night after night, remorselessly playing the odds as reflected in his credences. It should be noted that some accounts of the relationship between belief and credence do not permit this sort of independent variation between belief and credence, because they identify full belief with all and any credence that exceeds a certain threshold of confidence (e.g., over .7). I am committed to rejecting this sort of reductive understanding of belief in terms of credence.

There may be no exact historical precedent for this sort of account. I follow Thomas Aquinas in treating hope as an affective confidence, within the will. The theological virtue of hope, he says, is a confidence in achieving the ultimate end of human life, eternal union with God. But where does that confidence come from? Aquinas's answer, in contrast with what I am proposing, is that hope should always have its grounds on the cognitive side. In the theological case, hope is grounded in faith. He is too much an evidentialist to approve of hope—even divinely infused hope—without any cognitive underpinning. Accordingly, he thinks hope, if it is to be virtuous, must not exceed our cognitive expectations. Thus, our hope in our eternal reward must be balanced by a corresponding fear of divine punishment. Too much hope, unalloyed by fear, results in the vice of presumption (see Summa theol. 2a2ae qq. 18–19).

What I am recommending, in contrast, is epistemic presumption writ large: that we hope for, and so believe in, many things for which we have no rational expectation. This is different from believing on faith, because to have faith is not just to believe but also to have high expectations—a high credence in the likelihood of a thing's being so. Hope, as I am thinking of it, requires no such cognitive confidence. It begins from an optimism on the affective side—a cheerful willingness not to worry about the all-too-possible bad scenarios. This seems to me a more attractive doxastic stance than its fideistic counterpart, because it comes closer to adhering to the cardinal principle of evidentialism. The hopeful agent honors the evidence, through her credences, even while she takes liberties with her beliefs. Fideism requires a worrisome lack of integrity on the cognitive side, assigning probabilities in a way that the evidence does not warrant. The hopeful defeatist, in contrast, exploits the affective aspect of belief formation, by letting her sentiments lead her to belief, even while her credences reflect the depressing limits of our overall evidence.

In extolling such an attitude, I do not mean to suggest that it is within our voluntary control. Petrarch describes himself as too enmeshed in doubt to be hopeful, suggesting that it was not possible for him to see things otherwise. Hume makes it similarly clear that his conduct is governed by nature. Outside of his study, he just cannot help but believe things that, earlier, he had been capable of doubting. In this regard, ours certainly is the Humean predicament. Even if, on philosophical grounds, we find epistemic defeatism cogent, it would be almost impossible for most of us to cease believing the various commonsensical propositions about the world that we take ourselves to have good reason to doubt.

14. Is This Worth Hoping For?

After all this, one still may wonder: can such hopeful defeatism really count as knowledge? As should be clear by now, I take our preoccupation with the K-word to obscure these issues rather than to illuminate them. Still, even if ‘knowledge’ is just a generic honorific we apply to our beliefs, it is worth considering just how much honor there is in this outcome. Clearly, as in the case of faith, the question depends in part on how we constrain our hopefulness. Too much hope, or the wrong kind of hope, will strike everyone as absurd. Given the thesis of epistemic defeatism, the sorts of constraints available to us cannot be evidential. But that does not mean there can be no constraints, only that the constraints will ultimately be grounded in something other than objective evidence. Again it is helpful to draw a comparison to moral antirealism, where the denial of objective moral values does not put an end to moral theorizing. What shape theorizing might take in the epistemic domain, for the proponent of epistemic defeatism, is a question I have only just begun to ask.

Still, the question persists, can such hopeful beliefs count as knowledge? It seems to me not hard to imagine circumstances in which an affirmative answer is plausible. If, in particular, the world turns out to be close enough to how we hope that it is, and if the methods and mechanisms we have been relying on turn out to be well-calibrated to describing that world correctly, then ‘knowledge’ seems like a perfectly good term for describing what we have achieved. Speaking hopefully, we give ourselves the benefit of the doubt and use the word, even if we will never be in a position to confirm its applicability.

With such confirmation out of reach, our epistemic position is far from ideal. Indeed, in an obvious sense, our hopefulness is quite irrational. For if I recognize that the evidence points no more in one direction than another, then reason tells me that I really should be worried about my beliefs going wrong and so should not hazard belief. But such judgments about what we should and should not do need to be considered in light of the alternatives. From a narrowly epistemic perspective, our lack of good evidence should drive us toward Pyrrhonian suspension of belief. But in the larger scheme of things the rationality of that outcome looks highly questionable. Suspension of belief may be the best solution if the only thing in life we value is maximizing our ratio of true to false beliefs. In that case, given our impoverished epistemic state, the best move might be the limiting-case strategy of forming no beliefs. But inasmuch as we care about many things other than truth and falsity—inasmuch as we want to live rich, engaged lives—an attitude of hope recommends itself. We move ahead with life, hopefully, even while remembering the disappointing verdict of philosophical reflection on our epistemic predicament.

I have not argued that epistemic defeatism is true, only that it deserves to be taken seriously. We should take it seriously even though the prospect of its truth is hugely dispiriting. It would be much better if we could find good evidence for our beliefs. But philosophers who still tenaciously cling to that long-sought ideal should ask themselves whether their commitment itself is based in reason or is merely wishful thinking. We very badly want our beliefs to be ultimately grounded in good evidence, even if the long philosophical history of failure makes success look unlikely. At this level too, then, the choices are the same: you can have faith that epistemic defeatism is false, or you can merely hope. Once again, hope is better.