SORITES, ISSN 1135-1349
Issue #14 -- October 2002. Pp. 85-109
Copyright © by SORITES and Mark Walker
1.0 A Puzzle?
The general problem with justification is that the procedures we adopt, across all walks of epistemic life, appear highly permeable to difficulties posed by scepticism. The problem of justification is therefore in large part the problem of scepticism; which is precisely why discussion of scepticism is central to epistemology.Foot note 7_5
But at this level of abstraction, it might seem perplexing why the justification condition is singled out; for the traditional definition of knowledge is that it is equivalent to justified, true belief.Foot note 7_6 Given that the skeptic's goal is to undermine confidence in our epistemic claims, and that there are additional necessary conditions for knowledge, why is so much emphasis placed on the justification condition? After all, undermining any of the other conditions ought to work with equal effect. That is, unquestionably there is nothing about the logical relations among the various components of knowledge that may be used to differentiate them as intrinsically interesting for skeptics. If, for example, justification were the sole necessary condition for knowledge, and the other conditions figured in analyzes of sufficient conditions, then, the reason for attention to the justification condition might be apparent. However, given our present understanding of knowledge, then, necessarily, if any one of these necessary conditions is not satisfied then knowledge is impossible. Of course there may be reasons other than its logical relation to knowledge for focusing on justification -- perhaps there is some reason to suppose that justification is the weak link in the structure of knowledge. Whether this is so remains to be seen.
I mention this «puzzle» -- if it is indeed that -- only to illuminate the general strategy I hope to explore here. My argument will be that the same sorts of considerations that motivate skeptical questioning of the justification component of knowledge can be extended to the other definiens of knowledge. Our first task then is to examine the structure and definition of knowledge in order to shed light on other potential candidates for skeptical questioning. In section 3 I will attempt to articulate four types of skepticism based on this definition. Section 4 argues that these four types of skepticism can be supported by invoking counter-possibilities or «defeaters». Section 5 concludes with a brief look at some applications and consequences of this taxonomy of skepticism.
I should perhaps issue a caveat at this stage. If it is not already apparent this is a «big-picture» paper in the sense that it covers an enormous amount of area. Of necessity then exploration of many details of particular arguments will have to be sacrificed. I believe that, if my argument is persuasive at all, it is in the manner in which it unites what are perhaps the two largest epistemic enterprises of the late twentieth century-defining `knowledge' and responding to skepticism-into a single synoptic vision.
The proposal then is to analyze various forms of philosophical skepticism in terms of the necessary conditions of knowledge. However, an obvious obstacle stands in the way: Gettier's criticism of the traditional definition of knowledge seems to leave the concept of knowledge partially undefined. I will argue, however, that there is a definition of knowledge of sufficient clarity available -- at least for our purposes.
Let us, then, first look at the Gettier problem. The traditional definition of knowledge is often explicated along these lines:
S knows that p = Definition:
i) S believes that P
ii) It is true that P
iii) S is justified in believing that P.
Gettier offered a couple of counterexamples by way of refutation of the claim that these conditions are sufficient for knowledge. However, we will use an example adapted from Russell to illustrate this point.Foot note 7_7 Let us suppose that Fred has worked in an office for forty years and has observed a clock working reliably for the same period. One morning Fred glances at the clock and as a result believes that it is nine o'clock. What he does not know is that the clock stopped working exactly twelve hours ago. Thus, it is true that it is nine o'clock, Fred believes that it is nine o'clock, and Fred is justified in his belief that it is nine o'clock. However, the intuition of most is that Fred does not know that it is nine o'clock. Something seems to be missing.
Over the years we have seen a number of ingenious attempts to specify what this something missing is, the search for the elusive «fourth condition».Foot note 7_8 This is not our quarry. Rather, for our purposes, a much more modest proposal is sufficient. The leading idea is that a natural understanding of cases like that of the clock is that the agent has a justified, true belief, but the belief is not X, where `X' names precisely that awareness articulated by the Gettier type examples. That is, the concept X articulates the common understanding that is shared in response to these types of examples. Following in the great taxonomical tradition in biology we may name this concept X, `gettier'.
Assuming (as we shall) that the traditional analysis of knowledge is correct,Foot note 7_9 the following four conditions, then, are necessary and sufficient for knowledge:
S knows that p = DF
i) S believes that p
ii) It is true that p
iii) S is justified in believing that p.
iv) S's belief that p is gettier.
An obvious response to this definition is that it does not solve the Gettier problem, it (at best) names the problem. This criticism is correct if we are speaking about constructing a theory of knowledge, although not if we are attempting merely to make available a definition of knowledge. To see our way to this point notice what is implicit in what was said above: the traditional definition of knowledge is defined in terms of primitive concepts, not in terms of theories about these more primitive concepts. For example, it is not typically taken as a point against the traditional definition of knowledge that the condition «it is true that p» is neutral (or less charitably, ambiguous) among the competing theories of truth. Perhaps a full-blown theory of knowledge would require a full theoretical articulation of the constituent terms, in which case the debate over the nature of truth, e.g., correspondence versus coherence etc., would be to the point. But this is not what the traditional account sought. An exact parallel between the fourth condition and the other three would have us cite some «ready-made» concept. If the parallel holds we should be able to cite some primitive concept analogous to `truth', `justification' and `belief' to fill in the fourth spot. The actual literature on the fourth condition, however, looks more like the sorts of debates that take place when we are trying to articulate a full-blown theory, rather than merely citing some «ready-made» concept.
Naturally one might think that if there were such a ready-made concept, then it would have been a relatively simple matter to cite the concept, and the «Gettier problem» would have been solved long ago. The fact that rather complex theories for the fourth condition have been proposed, it might be argued, ought to be taken as some evidence that we lack such a concept. But do we actually lack such a ready-made concept? I think not. I propose that we see Gettier's accomplishment as pointing out that there is a concept that we have been able to understand and use, but not mention. It is this concept that allows us to identify Gettier-type cases and discuss various theoretical solutions to the Gettier problem, just as our concept of truth allows us to discuss various theories of truth. So naming this concept means we have a concept that we can now both use and mention, and it allows us to complete the task of defining knowledge. The project of providing a full-blown theory of the fourth condition remains as incomplete as before. Much of the literature on the Gettier problem, then, can be seen as a contribution to the task of articulating a theory. What is important to realize, however, is that there is no «Gettier problem» with respect to the definition of knowledge anymore than there is a problem with its sibling concepts, i.e., `truth', `justification', and `belief'. Whether this definition is acceptable, ultimately, will depend in large measure whether it proves useful. I hope to show that it can be put to good use below.Foot note 7_10
An analogy from the history of philosophy might help secure this point. Aristotle in the Nicomachean Ethics asks what the highest good attainable by actions is and asserts that most would say that it is `happiness'. As Aristotle points out, although this may answer one question it leaves unresolved the explication of happiness: some think that it is the possession of health, others say wealth or honor etc.Foot note 7_11 We can imagine that the concept of gettier plays a similar role: we may agree that knowledge has as one of its definiens the concept of gettier, but disagree when it comes to explicating what gettier is.
We ought to pause here to consider a serious objection to this definition, namely: that the concept of gettier is completely parasitic on that of knowledge. It is serious, since if this objection could be maintained, it would seem we have good reason for rejecting the proposed definition of knowledge. After all, we have an understanding of the other concepts used to define knowledge, independent of our understanding of knowledge, certainly it seems that we should demand this much of the gettier definien. This objection, however, is unsuccessful as the concept of gettier too has a life independent of knowledge; for we may ask whether a statement rather than a purported item of knowledge is gettier. Adjusting our previous example, let us suppose that Fred by some mental quirk (brought on by too many philosophy courses according to his friends) distrusts clocks. Fred thus does not know that it is nine o'clock, according to the traditional analysis, because he does not believe that it is nine o'clock. When Sally, Fred's boss, asks the time he replies that it is nine o'clock -- Fred realizes that his skepticism about clocks is not shared or appreciated by his boss. Fred's statement is justified -- as we have said the clock has worked flawlessly for forty years -- and the statement is true. Given that the clock stopped exactly twelve hours previous, it seems that Fred's statement is not gettier. The fact that we understand whether a statement is gettier or not independent of whether the statement is known shows that the concept of gettier is not completely parasitic on that of knowledge. The appropriate parallel here is how we can understand a statement being either true or justified independent of whether the statement is known. At least in this respect, the concept of `gettier' does not differ from its more illustrious sibling concepts.
In this section I will attempt to delineate four types of philosophical skepticism based on the four components of knowledge: justificatory skepticism, gettier skepticism, noetic skepticism, and aletheia skepticism. The conditions that justificatory and gettier skepticism are associated with are self-explanatory. Noetic skepticism concentrates on the belief condition of knowledge and aletheia skepticism questions the truth component.
As its name implies, justificatory skepticism focuses on the justification condition for knowledge. As will soon become apparent, justificatory skepticism is more or less coextensive with what is customarily considered the entirety of philosophical skepticism. Historically, there are two main ways in which this idea is commonly expounded: one is what has been referred to as the `regress problem' but, as Williams argues, is more aptly termed `Agrippa's trilemma'.Foot note 7_12 The other is what might be thought of as Cartesian skepticism. Let us take these in turn.
Agrippa's trilemma traces its historical roots to the Ancient Greeks who first noticed that skepticism seems a natural consequence of reflection on the process of justifying our epistemic claims. When confronted with a persistent skeptic it seems we are faced with a very unappealing trilemma. Suppose a justificatory skeptic asks us to defend some epistemic claim P. We might choose to defend it on the basis of some further claim Q, but the skeptic will then ask about the justification for Q itself. If we invoke R in support of Q then, short of some alternate strategy, we are faced with an infinite regress. The difficulty of course is that only two alternatives present themselves: either we invoke P again in our defense of R, but this invites the criticism of begging the question. The only alternative is to insist that some claims or beliefs are self-justifying, e.g., let us suppose that R justifies Q and Q justifies P, but R itself does not require any further justification, R is self-justifying. R, in other words, is an axiom. This suggestion at least has the correct form for silencing the skeptic. If we can get the skeptic to accept our chain of reasoning and the axiom then we will have met the challenge. The trouble of course is that this foundational response to skepticism has generally been considered a failure.Foot note 7_13 Roughly, the difficulty seems to be in discovering claims that are both specific enough to provide justification for statements like «I am sitting in front of a fire» and beyond skeptical doubt. Certain logical truths like «not every statement is true and false»Foot note 7_14 are perhaps beyond skeptical doubt, (or at least the best candidates for this role), but it is difficult to see how such abstract claims on their own could be used to support specific claims. On the other hand, if we turn to propositions that look like they might be more useful in supporting the fireside claim, e.g., a benevolent deity would not allow me to be radically deceived about such matters--these look like they are much more vulnerable to skeptical doubt. In any event, this is a thumbnail version of one understanding of the failure of the foundationalist answer to the skeptic.
The other form of justificatory skepticism is most famous and familiar from Descartes' evil demon thought experiment.Foot note 7_15 Descartes' point, to put it into a contemporary idiom, is that the evidence we have for our epistemic claims is often radically underdetermined, i.e., the evidence is compatible with (and perhaps equally supportive of) any number of hypotheses. As Descartes suggests in the Meditations, he has very good sensory evidence to suppose that he is sitting in front of a fire, but this evidence is also compatible with the hypothesis that a powerful evil demon is attempting to deceive him that he is sitting in front of a fire. A modern variant on this thought experiment--made popular in such movies as The Matrix--speculates that what appears real to us is merely a virtual world created by some supercomputer. The most gruesome versions of this thought experiment ask us to image that our brains have been surgically removed and placed in a vat of nurturing nutrients. The computer interfaces via implants designed to mimic the sorts of information that our sense organs would send along the nervous system. Thus, the evidence that I have for the belief that I am presently sitting at a desk writing this paper is also logically compatible with the suggestion that my brain is floating in a vat full of nutrients and a computer is generating for me the illusion that I am sitting at a desk. If we take such scenarios seriously then it seems that much of what we believe and claim to know might in fact be false.
It should be emphasized that this skeptical suggestion, that it is possible that all our beliefs (about the external world) might be false, is a point about justification. The argument from the underdetermination of evidence demonstrates that (standard) attempts to justify our epistemic claims typically do not imply the truth of the statements they are intended to justify. Take for example what we might think of as the `standard view of the world', which, for at least a certain segment of the world's population, consists in the conjunction of what Sellars termed the `manifest' and `scientific images', which correspond roughly to the beliefs and theories of common sense and contemporary science.Foot note 7_16 Cartesian skepticism says that our justification for believing the external world is more or less like we think it is in fact implies a disjunct along the lines of:
D: Either the world is more or less as standard view describes, or the brain-in-the-vat scenario is true, or the evil demon hypothesis is true, etc.
For as we have said, the evidence that we take to support our knowledge claims about the external world is logically compatible with the claim that we are deceived. Cartesian skepticism, then, can be seen as challenging us to eliminate the skeptical disjuncts of D. Of course this is precisely what the Cartesian skeptic suspects cannot be done, i.e., we have no way to justify our common sense and scientific view such that they are uniquely implied by our justificatory practices. That is, the Cartesian argues that there does not appear to by any a priori reason why we should favor the standard view, nor does the empirical evidence provided by our senses allow us to favor the standard view. Whether we ought to accept the Cartesians' assessment of the epistemic terrain remains controversial. (I discuss (but do not defend) the Cartesian view in more detail in section 4.1).
The conclusion of both types of justificatory skepticism is that the ability to justify putative items of knowledge is beyond our abilities, i.e., the justification condition cannot be satisfied, and hence, knowledge is impossible.Foot note 7_17 I hope that this brief exposition is sufficient to indicate that these two types or families of skepticism are more or less coextensive with what is normally considered philosophical skepticism.Foot note 7_18 I say `families' because it is clear that other famous skeptical hypotheses, e.g., Russell's suggestion that it is difficult to prove that the world wasn't created only five minutes ago, skepticism about induction, or skepticism about other minds, have the same general form as Cartesian skepticism.
The first nonstandard philosophical skepticism that we should examine is gettier skepticism.Foot note 7_19 Gettier-type examples typically involve very local epistemic claims, as we saw in the clock example above. The gettier skeptic asserts that it is quite possible to imagine much more radical thought experiments, as the `Ed the lucky brain-in-a-vat' example demonstrates. Let us suppose that Ed is the hapless victim of a «brain kidnapping». The proverbial mad scientist has surgically removed his brain and placed it in a vat. The mad scientist has set-up the program such that in the morning when Ed awakes Ed will believe that he is reading his copy of the «Sydney Morning Herald» when in fact he will be reading the scientist's vat-image imitation. The scientist plans to plant all sorts of implausible stories to test Ed's gullibility, e.g., that President Bush is a (closet) card-carrying Trotskyite, that scientists have discovered that the moon is actually made of cheese (other than a thin surface crust of rock). However, that night there is a mammoth electrical storm and a bolt of lightning hits the computer controlling Ed's «vat world». By massive coincidence, the computer software is scrambled such that the program will in fact provide Ed with a lot more truths than the scientist had intended. It turns out that the vat-image imitation of the morning paper is word for word identical to the real «Sydney Morning Herald». Ed might read for instance that the stock market is down and thus believe the truth that the stock market is down, when in fact the stock market is down. The computer's software might engender the belief in Ed that it is a cloudy day, and it is in fact a cloudy day, and so on. We might suppose that Ed has as many true beliefs about the world (which do not involve him) as he would have if he had not been kidnapped. Ed's beliefs about the current state of the world -- the stock market, the weather etc. -- even though they are true and justified, fail to qualify as knowledge. His beliefs are not gettier. They depend on the most unlikely of circumstances, e.g., the lightning hitting the computer in just a certain way. Thus, a skeptic may grant that a brain-in-a-vat might have true, justified beliefs, but deny that such a brain would in fact know. For the brain's beliefs might not be gettier. Gettier skepticism is a form of philosophical skepticism then because it advises that knowledge is impossible because we have no reason to suppose that the gettier condition of knowledge is satisfied.
Noetic skepticism differs from the previous skepticisms in that it questions whether we are capable of formulating the appropriate types of belief in the first place -- never mind the subsidiary question of whether these beliefs are true or gettier. It claims that the hypothesis that correctly describes the truth might be beyond the «reach of our minds» -- to use Nagel's formulation.Foot note 7_20 The idea that we are conceptually limited is one familiar from the history of philosophy and can be traced back at least as far as Heraclitus:
(Fragment 83) The wisest man will appear an ape in relation to God, both in wisdom and beauty and everything else.
(Fragment 79) Man is called childish compared with divinity, just as a boy compared with man.Foot note 7_21
This line of thought has recently been recast within a naturalistic framework. Jerry Fodor writes:
...so long as the class of accessible concepts is endogenously constrained, there will be thoughts that we are unequipped to think. And, so far, nobody has been able to devise an account of the ontogeny of concepts which does not imply such endogenous constraints. This conclusion may seem less unbearably depressing if one considers that it is one which we unhesitatingly accept for every other species. One would presumably not be impressed by a priori arguments intended to prove (e.g.) that the true science must be accessible to spiders.Foot note 7_22
Heraclitus' analogies and Fodor's reflections on the epistemological consequences of evolutionary theory are suggestive of a quite radical form of skepticism: namely, that we are congenitally incapable of knowing very much of anything.Foot note 7_23 There are two ways of articulating noetic skepticism: One is what I shall term the `proper subset argument', the other is the `Kantian argument'.
This formulation of the argument takes its cue from Heraclitus' ontogenetic analogy: human understanding is said to stand to an omniscient understanding as a child's understanding stands to ours. Obviously the younger we imagine the child in the comparison the less flattering it looks to our level of intelligence and knowledge. Plausibly, the average child of 15 is capable of understanding most of what an average adult knows; so let us suppose, for illustrative purposes only, that a 15 year old can understand and know 80% of an average adult's knowledge. This comparison might arguably be said to generate merely a local skepticism when applied to our own case. For if we stand to an omniscient being as the 15 year old stands to us, then we have the conceptual capacity to understand 80 % of an omniscient being's knowledge. However, if the comparison is between an adult and a child of 2, and the average child of 2 knows (let us suppose) 20 % of what the average adult knows, then we have the basis for generating a global skepticism. For now we must imagine that it is possible for us to know only 20 % of what the omniscient being knows, that is, 80 % of the omniscient being's knowledge is unknowable by humans. The discrepancy seems at least as large if we switch to Heraclitus' phylogenetic analogy, for it seems quite likely that apes know much less than 20% of what a normal human adult knows. How far we can push this analogy is an interesting question. Stephan Körner suggests that our level of intelligence compared to a superintelligence might be like that of a worm's to a human.Foot note 7_24 Worms are almost entirely ignorant about the nature of the universe so what values should we set here, or does the analogy become unintelligible at this point? It will take us too far a field to investigate these questions. In any event, to the extent that we accept the proper subset argument, it seems in harmony with the philosophical skeptic's recommendation for epistemic modesty.
A second line of argument derives from Kant who developed a version of noetic skepticism about the external world considered as a thing in itself. This is not the place to explore the subtleties of Kant's position; we will have to be content with a thumbnail version of his noetic skepticism: in this connection the upshot of Kant's argument of the first Critique is that some of the most basic features of our conceptual scheme may serve to distort the nature of things in themselves.Foot note 7_25 For example, one of Kant's suggestions is that our space/time intuitions may misrepresent the true nature of things in themselves, for we cannot rule out the possibility that God might intuit the world in a different manner, i.e., we have to contend with the possibility that He might not possess our sensuous form of intuition, but instead, work with an intellectual intuition. Kant also seems to think that God knows the world directly with his intellectual intuition, He does not have a discursive understanding like us since: «...all his [God's] knowledge must be intuition, and not thought, which always involves limitation.»Foot note 7_26 Kant argues that we cannot say much about how this form of intuition might work; it seems to be beyond our conceptual grasp, indeed, he says that we can say nothing positive about this alternate means of intuiting the world.Foot note 7_27 Can we do any better? Not much, perhaps, for as with the proper subset argument, it seems that we must resort to analogies to clarify this skepticism, for obviously we cannot describe in detail that which lies beyond the limits of our thought. An analogy that might be of some use is that of a map, a two dimensional representation of our planet, versus a three dimensional representation like a globe. It is well-known that many standard maps tend to distort somewhat the exact spatial relations among the parts of our planet, e.g., Greenland often tends to get represented as much larger than it really is. Not only are maps usually inherently less than perfectly accurate they are also incomplete, e.g., the shortest distance between any two points on the planet will almost always yield the wrong answer because maps ignore the three dimensional topography of our earth, e.g., the shortest distance to China from Canada, as many children digging in their backyard know, is through the earth. To apply this analogy to our cognitive situation we might image that our understanding of the external world (in itself) is like that of a map, whereas an omniscient being's is more like that of a globe. If this is the case then we can see how our understanding is both conceptually capped -- we fail to understand certain concepts just as a map fails to represent the three dimensional aspects of the world -- and it may be distorting, just as maps regularly represent parts of the planet as either too large or too small. Notice too that our analogy helps us understand how this form of noetic skepticism can be explained in a manner consistent with naturalism. We may suppose that the additional information of a «globe» type representation of reality -- assuming that it is even available in the gene pool -- may have conferred little or no additional survival value to our ancestors while taking up cognitive resources. That is, maps may be less than perfectly accurate and conceptually incomplete, but they can be extremely useful. If this analogy holds, then, we can make sense of the idea that the true representation of the external world is beyond our cognitive capacity, and furthermore, that our form of representing the external world may be false, but nevertheless, our form of representing the external world may be extremely useful.
Aletheia skepticism focuses on the truth condition for knowledge: it says that truth is (for the most part) beyond our ability to grasp.Foot note 7_28 F. H. Bradley is perhaps the most illustrious inspiration for aletheia skepticism.Foot note 7_29 Bradley argued that «There is no possible relational scheme which in my view in the end will be truth...I had long ago made it clear (so I thought) that for me no truth was quite true...»Foot note 7_30 In other words, all human attempts to ascertain the truth fail. If we accept this thesis, then from our definition of knowledge, philosophical skepticism follows as a consequence. If one agrees with Bradley then skepticism follows, for one of the necessary conditions for knowledge is unattainable by humans. Of course this begs the self-referential question: is a Bradley inspired aletheia skepticism itself true? Presumably Bradley would be forced to say that it is not quite true. It would probably be premature for the anti-skeptic to celebrate, since it does not follow that anti-skepticism has ascertained the truth on this point either, for Bradley claims that no truth is quite true. Bradley had a doctrine of degrees of truth that might help here. Specifically, while Bradley claims that no human attempts to ascertain the truth can ever be quite true, nevertheless, he argues that some claims are closer to the truth than others. Thus, it may be that aletheia skepticism is closer to the truth than anti-skepticism.
Bradley's argument for the claim that «no truth is quite true», in outline, is that human cognition is shot-through with contradictions, but reality in itself is not contradictory; rather reality is a self-consistent whole:
Truth is an ideal expression of the Universe, at once coherent and comprehensive. It must not conflict with itself, and there must be no suggestion which fails to fall inside it. Perfect truth in short must realize the idea of a systematic whole.Foot note 7_31
Hence, it follows, according to Bradley, that humans are incapable of grasping the truth, which, as we have noted, implies philosophical skepticism.
Chomsky suggests a more naturalistic route to aletheia skepticism, he writes,
What is the relation between the class of humanly accessible theories and the class of true theories? It is possible that the intersection of these classes is quite small, that few true theories are accessible. There is no evolutionary argument to the contrary. Nor is there any reason to accept the traditional doctrine, as expressed by Descartes, that human reason is a «universal instrument which can serve for all contingencies.» Rather, it is a specific biological system, with its potentialities and associated limitations. It may turn out to have been a lucky accident that the intersection is not null. There is no particular reason to suppose that the science-forming capacities of humans or their mathematical abilities permit them to conceive of theories approximating the truth in every (or any) domain, or to gain insight into the laws of nature.Foot note 7_32
The general strategy of the aletheia philosophical skeptic is the same as the other types of philosophical skepticism: the argument is that knowledge is impossible because one of the necessary conditions for knowledge, viz., truth, is beyond the reach of humans.
Before moving on we should dwell for a moment on the question of the relation between noetic and justificatory skepticism. It might be objected that these two types of skepticism are not as distinct as advertised: both ultimately, it seems, want to conclude that formulating true beliefs about reality is impossible for humans. Certainly there is something to this objection, given that it seems that we might rightly hope that a full philosophical account of `belief' and `truth' will speak to how the concepts intersect. But even if one believes with the view that there are strong conceptual linkages between belief and truth,Foot note 7_33 still there may be some point to distinguishing these two skepticisms; if for no other reason than to mark the strategy they used to reach the conclusion that true beliefs about reality are impossible for humans. As we saw in the discussion of noetic skepticism, Fodor approaches the conclusion that true belief is impossible on the basis of the idea that there are concepts and thus beliefs that we cannot formulate. Chomsky, on the other hand, in the quote above, emphasizes that it is a true theory that is beyond our grasp. We can emphasize the divergence here by noting that for the noetic skeptic it does not follow that truth is beyond our grasp: it is logically consistent with noetic skepticism that we have some non-belief based access to the truth, e.g., imagine that someone believed that art provides a «feeling» or «intuition» about the true nature of reality, but that humans are not capable of formulating this access to truth in terms of belief.Foot note 7_34 On the other hand, this non-belief access to truth is incompatible with aletheia skepticism for it is the doctrine that the truth about reality is beyond our grasp; not simply that true belief is beyond our grasp. Conversely, noetic skepticism is the thesis that reality transcends our ability to believe it, so it follows that true belief is beyond our grasp. But so too is false belief. Our situation, if the noetic skeptic is correct, is like that of a monkey's vis-à-vis physics: it is not that we think that monkeys have a false view about physics -- that they are Newtonians when they should be Einsteinians -- rather they do not have physics-type beliefs at all. A noetic skeptic, in contrast, can allow that we have beliefs in the relevant domain, it is simply that these beliefs are false. All this demonstrates is that noetic skepticism and aletheia skepticism are logically distinct. As intimated, in practice we might suppose that the sorts of considerations that motivate noetic and aletheia skeptics will be very similar, e.g., the sorts of considerations that Fodor and Chomsky raise are not entirely dissimilar.
We have had occasion to note that justificatory skepticism is typically seen as co-extensive with philosophical skepticism itself. And although I have delineated three other types of skepticism, naturally this in itself does not show that these other skepticisms are viable or indeed philosophically interesting in their own right. In other words, nothing said thus far suggests that skeptics are not right to focus on the justification condition, for it may be the weak link. However, in this section I will argue that the same considerations used to motivate Cartesian justificatory skepticism can be employed with equal effect by the other three types of skepticism. Specifically, Cartesian skeptics suggest that we can conceive of alternatives to our epistemic claims that act as «defeaters» to the knowledge claims. Our first task then will be to consider the general form of the defeater stratagem and then apply it to the other three skepticisms.
The first step of the argument is to extend the analysis of Cartesian skepticism developed in section 3.12. A. C. Grayling suggests this formulation of the formal aspect of Cartesian skeptic's underdetermination argument:
Stated succinctly and formally, scepticism is the observation that there is nothing contradictory in the conjunction of statements s embodying our best grounds for a given belief p and the falsity of p.Foot note 7_35
As we shall see, Graylings formulation (at best) characterizes justificatory skepticism. Confining our attention for the moment to justification we may nonetheless wonder why this observation should present a problem for our ordinary justificatory practices, unless one assumed some sort of (deductive) foundational model of justification. Such a foundational model would insist that the only adequate form of justification is one where P can be deduced from some first principles. Such a form of justification would rule out the sort of skepticism Grayling is speaking of, but of course such a view is considered by most to be far too demanding. On the other hand, if we accept this model then it seems we must adopt justificatory skepticism; given, as we noted above, that very few statements (if any) meet this stringent requirement. However, as Stroud and others have argued, this view may actually mistake cause and effect, i.e., foundationalism is an attempt to answer skepticism rather than a cause of skepticism itself.Foot note 7_36
In fact, Stroud's understanding of Cartesian skepticism suggests that it seems to follow as a natural extension of our everyday justificatory practices. Take a simple example. Suppose Anna asks, «Does anyone know where Dawn is?» and Barry answers, «Yes, she is at the bar». Our normal justificatory practices suggest that if Barry does in fact know that Dawn is at the bar then he has ruled out potential defeaters to his knowledge claim (or is in a position to rule out such defeaters). For example, if Anna says, «Are you sure? I know Dawn is dead-set against drinking, but her (identical) twin sister Kimberly is a barfly if there ever was one.» Let us suppose that Barry is not in a position to eliminate this alternative, (he did not know that Dawn had a twin sister); it seems quite natural to say that Barry does not know that Dawn is at the bar. So in this case we seem to have a very straightforward version of an extremely local version of skepticism. (Such skepticism is of course not a type of philosophical skepticism since its scope is too narrow).Foot note 7_37 More formally, the skeptical result seems to be generated by the following line of thought:
C: If Barry knows that Dawn is at the bar then Barry knows that it was not Kimberly (Dawn's twin) that he saw at the bar.
Since Barry is not in a position to rule out the relevant defeater -- that it was Kimberly that he saw -- Barry does not know that Dawn is at the bar. We can see then that this sort of local skepticism fits the model suggested by Grayling: Barry's evidence for the claim that Dawn is at the bar is consistent with it being false that Dawn is at the bar. If Barry is to legitimately claim to know that Dawn is at the bar then it seems he must be in a position to reject the defeater, that is, it seems that it is a condition of Barry's knowing that Dawn is at the bar is that he knows that the person he saw there was not Kimberly.
More controversial of course is whether a global skepticism can be generated in the same manner. Formally at least, the Cartesian philosophical skeptic's view is identical to the local skepticism sketched above. Indeed, there is nothing that would prohibit using Cartesian-type thought experiments to underwrite a local skepticism:
C': If Barry knows that Dawn is at the bar then Barry knows that he is not a brain-in-a-vat.
For if Barry does not know that he is not a brain-in-a-vat then he cannot rule out the defeater that what he thinks he saw was in fact merely a computer generated image of a person, i.e., his evidence has nothing to do with whether there was a real person named Dawn at a real bar. Of course the Cartesian skeptic will argue that Barry does not know that he is not a brain-in-a-vat. In any event, if we accept C' as a natural extension of our everyday justificatory practices, and we despair of showing that Barry knows that he is not a brain-in-a-vat, then it follows that Barry is in no position to claim that he knows that Dawn is at the bar.Foot note 7_38 Obviously, whether Cartesian-type defeaters demonstrate that what we do not know is a controversial matter -- by `Cartesian-type defeaters I mean thought experiments like the brain-in-a-vat or the evil demon. What is clear is that if it is permissible to invoke such defeaters then they can easily be used to generate a global form of skepticism:
C1: If we know that the standard view is true in the main then we must know that Cartesian-type defeaters are false.
If one of these Cartesian-type defeaters is true, if, for example, we are brains-in-a-vat, then our claims to know the manifest and scientific image would be undermined: our knowledge claims would at best be about a computer generated world rather than the world outside of the macabre experiment that we are unwilling participants in. Cartesian-type defeaters are those skeptical cases where the truth, belief, and gettier components of knowledge are not the subject of skeptical challenge but simply the justification component.
Clearly there are two broad strategies that one might use against the skeptic: one can accept the conditional and show that the skeptic's rejection of the consequent is mistaken, or one can deny the conditional. With respect to the former, one possibility would be to construct a transcendental argument along the lines suggested by Putnam to argue that we know we are not brains-in-a-vat.Foot note 7_39 One way to deny the conditional is to follow contextualists who argue (roughly) that to insist that we eliminate Cartesian-type defeaters is to raise the standards beyond what is accepted by our everyday justificatory practices.Foot note 7_40 Whether this defeater stratagem is viable remains controversial within philosophy; and it is certainly not my intention to enter into this debate. Rather, let me emphasize that the rapid sketch I have provided here is simply meant to describe the defeater stratagem, not to defend it. What I will defend is the claim that, if we accept the defeater stratagem, it can be employed by the other three types of skepticism.
Let us consider then how the defeater stratagem can be applied to gettier skepticism. To emphasize how this differs from Cartesian skepticism recall Grayling's formulation of Cartesian skepticism: there is nothing contradictory in the conjunction of statements: s embodies our best grounds for a given belief p, and the falsity of p. There are two distinct reasons to reject Grayling's formulation; assuming that it is correct to interpret it as a necessary condition for skepticism. First, as we saw above in the `Ed the lucky brain-in-a-vat' example, it is quite possible to formulate a global form of skepticism that does not invoke the possibility that our epistemic claims are false. After all, in the case of the kidnapping of Ed we supposed that his beliefs were true even after he had been envatted. Thus, this case does not rely on the observation that our grounds for P are consistent with the falsity of P. Notice too that this example is a type of defeater: if we know that the standard view is true in the main then we must be in a position to eliminate the possibility that we are in exactly the sort of situation described in Ed's case.
A second means to demonstrate that gettier skepticism is a viable independent form of philosophical skepticism is to contemplate the example of the `Brothers Evil'. Descartes imagined a single evil demon; let us suppose that there are two demons, brothers as it turns out. They bickered over how to deceive people until the Ancient Greek civilization gave them the idea to divide their territories up into Thought and Being. One brother is in charge of manipulating the thoughts and beliefs of people, the other manipulates the world. Twice the error -- so they thought. Now imagine that the hapless humans of this example reach a point of ideal justification of the manifest and scientific image, i.e., they have reached what is sometimes referred to as the Piercean end of inquiry. The day of the party celebrating this epistemic victory, brother Thought surgically removes everyone's brain and places them in a vat and hooks up a program intended to radically deceive. One of the beliefs he engenders is that all the bodies of the solar system are made out of different cheeses, the Moon is feta, Mars cheddar and so on. Brother Being working independently had his own plans. (The two brothers had not talked since the publication of The Critique of Pure Reason when brother Thought taunted Brother Being by calling him «mere noumena»). Quite independently brother Being had decided to radically rearrange the universe in order to mock the end of enquiry celebrations. As fate in philosophy examples would have it, one of the first things he did was to turn all the bodies of the solar system into cheese, the moon was turned into feta, Mars cheddar, and so on. Of course it turns out that all the bizarre beliefs about the world that brother Thought engendered were exactly matched by bizarre changes in the world orchestrated by brother Being.
To ensure that considerations of truth and justification have been removed let us suppose that we endorse a version of anti-realism that says, necessarily, if our theory is ideally justified then it is true. This form of anti-realism would then maintain that there is a contradiction in the conjunction of statements s embodying our best grounds for a given belief p and the falsity of p. This in itself would demonstrate that Grayling's understanding of skepticism has been refuted in the brothers Evil example. For having reached the end of inquiry there is no room, according to this version of anti-realism, to suppose that our ideally justified theory is not true. If one accepts Grayling's formulation as a necessary condition for philosophical skepticism then this form of anti-realism has defeated philosophical skepticism. Yet if we accept there are alternate formulations of skepticism then this is not the end of the matter. For the philosophical skeptic still has the «fall back» position of saying that, although our ideally justified theory is true and believed by us, nevertheless, we may not know this theory. The reason of course is that the theory may not be gettier, as the above example demonstrates. The skeptic in effect says: «granted that given anti-realism, our ideal justification leads to truth no matter what the outcome of events, nevertheless, there are better and worse ways this could happen. It would be better if our justification connected in a way that was gettier rather than non-gettier.» It is perhaps worth noting here that the gettier skeptic is able to drive a wedge between justified true belief, on the one hand, and knowledge on the other, because what is meant by `ideally justified' here is the sort of justification humans might reasonably be expected to amass. To insist that by `ideally justified' we mean that all the logical possibilities have been eliminated is to insist on a standard of justification that looks unattainable by humans, for it would seem to presuppose that we have a «God's eye view» of the universe. Anti-realists have generally wanted to provide epistemology with a «human face».Foot note 7_41 It is precisely because of this relaxed standard -- i.e., not insisting that an ideal theory must eliminate all logical possibilities -- that gettier skepticism is able to get a toehold. From the anti-realist's perspective, without this «humanizing» move -- i.e., constructing epistemological theories that acknowledge our finitude -- we will forever be mired in an unproductive skepticism. What the brothers Evil example shows is if anti-realists hope to eliminate skepticism then they will need to do more than assert that our ideal theory must be true, they must assert that it is also gettier. Perhaps the easiest way for a Piercean anti-realist to achieve this is to stipulate that at the end enquiry our beliefs are not only true and justified they are also gettier. This would guarantee that we have knowledge and not just justified true belief. Without this stipulation there is still a potential gap between justified true belief and knowledge that the skeptic might easily exploit.
Gettier skepticism, then, can be motivated by a natural extension of the Cartesian's strategy to explore defeaters as a means to undermine knowledge: we have seen how it is possible to construct an alternative or defeater to one anti-skeptical strategy that does not require us to question whether the knowledge claim is true or justified. The gettier skeptic, then, maintains that
C2: If we know that the manifest and scientific images are true in the main then we must know that gettier-type defeaters are false.
Gettier-type defeaters are those skeptical cases where the truth, belief, and justification for a putative knowledge claim are not in question; rather, the question is one of whether the putative knowledge claim is gettier. The conclusion to draw is that gettier skepticism is independent of Cartesian skepticism and that it can be supported using the same defeater stratagem.
The defeaters stratagem can also be used in support of noetic skepticism, indeed, I think it can be shown that noetic skepticism is epistemologically prior to Cartesian skepticism in the same way that Cartesian skepticism is epistemologically prior to the epistemological stance of common sense and science. I will argue that it may be a very real possibility that we cannot eliminate defeaters that we cannot conceive.
The first step in this argument is to consider how Cartesian skepticism understands the realm of alternative hypotheses or defeaters. A natural interpretation of Cartesian skepticism is that the class of defeaters is extremely broad; any alternative that is conceivable seems to be up for consideration. Clearly the realm of conceivable alternatives is larger and more encompassing than that which is associated with the epistemological point of view of everyday life. Indeed, as our discussion above indicates, in effect the Cartesian skeptic's point is that the realm of defeaters countenanced by our everyday justificatory practices are not sufficiently expansive in their scope.
But is the realm of conceivability the appropriate one, that is, is it expansive enough to include all potential defeaters? There are, I believe, good reasons to think that this realm is also too restrictive. Consider the following sort of example: suppose that there are just two contested hypotheses about the nature of the universe. The standard model, or big-bang theory, H1, describes the universe as evolving from some primordial singularity. The steady state hypothesis, H2, proposes that the universe has always been more or less as it is now. Of course the steady state hypothesis has been rejected by contemporary science; but we can imagine a Cartesian skeptic arguing that we do not know that H1 is true because there is at least a conceptual possibility that we are in error. While our best available empirical evidence supports H1, it is logically possible that H2 is true. We might be asked by a skeptic to imagine an evil demon has arranged all sorts of false clues; e.g., the «alleged» background radiation of the universe leftover from the big bang was simply planted there by the epistemic fiend in an attempt to mislead us. The Cartesian skeptic does not suggest that we are ignorant of the conceptual alternatives, for they allow that we might entertain the possibility that H2 is true. A noetic skeptic could agree that H1 and H2 describe the only two hypotheses about the universe that are worthy of human scientific scrutiny but this does not rule out the possibility that the «complexity» hypothesis H3 is true. It suggests that the theory that best describes the universe must posit a billion billion billion billion billion billion initial conditions, and each of these initial conditions requires at least the same number of bits of information to describe it. Such a hypothesis, let us suppose, is far too complex for any human to conceive, yet it is logically incompatible with the truth of H1 or H2. The noetic skeptic then argues that the possibility of H3 demonstrates that we might forever be ignorant about the truth of the universe. So just as the Cartesian skeptic says that everyday epistemic practices are not sufficiently critical about the scope of relevant alternatives, so the noetic skeptic says that the Cartesian skeptics are not sufficiently critical in their evaluation of the relevant alternatives or defeaters. We would need some guarantee that we are in a position to conceive of the relevant alternatives. Perhaps the Cartesian would have been better off to tackle the problem head-on by appealing to the realm of what is logically possible. However, this option suffers from the mirror imagine of the previous. For if the Cartesian says that the alternatives that are to be rejected are all those that are logically compatible with the evidence then we need some assurance that we are in fact capable of surveying these potential defeaters. The sort of possibility raised by H3 suggests that we may lack the cognitive wherewithal to appreciate in sufficient detail every theory in this set.
There is an obvious and potentially decisive objection here: there is no straightforward way of describing the sorts of defeaters that might lie beyond human abilities to cogitate. After all, if they could be described then a fortiori they would not be beyond human ability to cogitate. In comparison with Cartesian skepticism, noetic skepticism suffers a serious disadvantage in that it is not able to describe in any detail the potential defeaters. Clearly part of the power of Cartesian skepticism is precisely the fact that the sorts of thought experiments Descartes engages in are easy to conceive. Yet to insist on a detailed understanding is to misunderstand noetic skepticisms, for it is precisely the possibility that we cannot grasp all the relevant possibilities that underlies noetic skepticism. While the noetic skeptic cannot describe in detail the sorts of possibilities that are conjectured to lie beyond our ability to conceptualize, noetic skeptics are not without resources. It is possible for us to see how others failed to conceive of relevant defeaters. One famous example is that historically it seemed reasonable to claim to know that Euclidean geometry best describes the world. A skeptic in the eighteenth century might have pointed out against this position that we cannot rule out the possibility that there are alternate geometrical configurations of the universe that are beyond our ability to conceptualize in any detail.Foot note 7_42 Of course this skeptic would have been at a disadvantage in that she could not describe in any detail the content of these alternate geometries, although, as we now know, subsequent developments in geometry (and physics) proved the skeptic right. In other words, from our epistemically advantaged point we can see that the possibility of such defeaters was indeed correct, given that we now accept a non-Euclidean geometry as the geometry of the universe. It seems we are in no position to eliminate the possibility that some advance civilization stands to us in a similarly epistemically advantaged position. Furthermore, it seems that we are in no position to suppose that there are not congenital limitations to the sorts of possibilities that we can conceive in any detail, just as the realm of possibilities that children and animals are limited from our perspective. A monkey, it seems, cannot even conceive of the thought experiment that it might be a brain-in-a-vat nor perhaps can the average (human) three year old. Given our finitude, it seems that we have every reason to suppose that there may be any number of defeaters beyond our ability to cognize just as there are any number beyond the comprehension of animals and children.
This problem is only exacerbated when we think in terms of how the evil demon or the mad scientist from the brain-in-a-vat scenario might be used in support of noetic skepticism. The mad scientist, for example, is often thought to interface with the kidnapped brain at the point where a brain would normally receive information input from the nervous system. The nerve endings that would normally extend to our sense organs and other parts of our body are said to receive artificial input directed by a supercomputer. But suppose the evil demon or mad scientist took the deception one step further and actually «reprogrammed» our reason such that we have a «blind-spot» in our reason for precisely those concepts and beliefs that describe our actual world. If we allow the possibility that we have purposely been redesigned by some malicious agent so as not to be able to conceive the world then we have a defeater that seems impossible (for humans) to defeat.Foot note 7_43
The Cartesian skeptic of course says that the justificatory practices of everyday life are not sufficiently critical, that we can conceive of all sorts of defeaters that lie beyond the realm considered relevant to everyday justificatory practices. Likewise, noetic skeptics believe that the realm considered by the Cartesian skeptic is not sufficiently critical, that there may be a number of defeaters that lie beyond our ability to conceive.
The discussion of aletheia skepticism here can be brief because the argumentative structure exactly parallels that of the noetic skeptic: the aletheia skeptic suggests that there are truths beyond our grasp and we cannot rule out the possibility that a defeater lies within this set. Again, the Cartesian skeptic's position does not look sufficiently critical given that we want to pursue the defeater stratagem vigorously. To see this recall the conditional used to support Cartesian skepticism:
C1: If we know that the standard view is true in the main then we must know that Cartesian-type defeaters are false.
The aletheia skepticism suggested by Bradley's claim that «no truth was quite true», points toward the following conditional
C4: If we know that the standard view is true in the main then we know that aletheia-type defeaters are false.
By `aletheia-type defeaters' I mean of course the sorts of possibilities that Bradley and Chomsky raise: that we may have no or little access to the truth. It seems appropriate then to draw the following conditional conclusion: If the defeater stratagem is viable, and if Bradley's aletheia skepticism is correct (has a higher degree of truth than its negation) then knowledge of the standard view is impossible.
I have attempted to make two points in this paper: First, that a full taxonomy of philosophical skepticism shows that there are alternatives to Cartesian skepticism. Second, if viable, the defeater stratagem can be used to support these alternative skepticisms. Both of these points (obviously) could bear further scrutiny than is possible here. With respect to the former it might prove profitable to examine the historical antecedents of these skepticisms. We noted briefly a possible connection between Kant and Bradley with noetic and aletheia skepticism respectively. Descartes is of course connected with `Cartesian skepticism', but even Descartes at times seems to have noetic skepticism in mind, although this is not often noted.Foot note 7_44
Another avenue to explore is how contemporary answers to skepticism figure into this classification. I will mention three, in very broad and rapid brush strokes, to provide some idea of the fecundity of this taxonomy. One example is how causal theories of content or belief are thought to answer skepticism about whether we know that we are not brains-in-a-vat.Foot note 7_45 One anti-skeptical view put forward is, very roughly, that there is a contradiction in saying that `I am a brain in a vat'. For if I am a brain in a vat then `vat' means in my language something like a type of computer generated image; but I am not a brain in a computer generated image. Many, I think, have found puzzling the idea that the argument that says that, if I am a brain-in-a-vat then I cannot think that I am a brain-in-a-vat, is an anti-skeptical argument.Foot note 7_46 Certainly this puzzle remains even if one accepts that my utterance of `I am a brain-in-a-vat' is false. One possibility to explore is that, even if this response is successful against the Cartesian skeptic, one may wonder how robust it is against the noetic skeptic. On this understanding the Cartesian skeptic might be seen as saying that one can conceive of the defeater that `I am a brain-in-a-vat'. The causal theory of meaning or belief refutes this suggestion by saying that this statement must be interpreted as false. On the other hand, this response to the Cartesian skeptic seems to bring to the fore noetic skepticism: if my beliefs or meanings are circumscribed by the realm of what I causally interact with then it seems there are any number of thoughts and beliefs that I am unequipped to think -- including a detailed picture about how I am being deceived. If one accepts this taxonomy of skepticism it suggests a line of defense for the philosophical skeptic against such arguments.
Another example is John McDowell's discussion of skepticism in Mind and World. McDowell starts out looking like he is going to discuss noetic skepticism, but as the discussion proceeds it is clear that he has justificatory skepticism in mind.Foot note 7_47 This is perhaps most evident in part of his response to skepticism. One of the problems he struggles with is the idea that we share with animals sense organs that are (in some sense) equivalent in their information gathering abilities, for example, other creatures besides humans have stereoscopic color vision. A crucial difference is that we use the information supplied by our sense organs to justify knowledge claims whereas animals do not. One question McDowell investigates is how this is possible. His answer to this is quite complex, but for our purposes we can concentrate on his proposal (inspired by Aristotle and Kant) that we have a «second» nature that is fused with our animal nature. With this taxonomy of skepticism we can see that, even if McDowell has formulated a reply to the Cartesian skeptic, he has not answered the noetic or aletheia skeptic. For they will wonder about the abilities of this «second» nature vis-à-vis other possible knowers. How do we know, for example, that in addition to an animal and second nature the Alpha Centaurians do not have a «third» nature and the Beta Centaurians a «fourth» nature and so on that eclipses our nature in the same manner that our «second» nature eclipses that of animals? These sorts of questions are not addressed by McDowell. Our taxonomy of philosophical skepticism suggests that they ought to be if the answer to the skeptic is to be complete.
A more complete answer to skepticism can be found in the work of Donald Davidson.Foot note 7_48 Davidson discusses what we are calling `justificatory skepticism' most notably in «The Method of Truth in Metaphysics» and «A Coherence Theory of Truth and Knowledge».Foot note 7_49 Although he does not discuss it under the rubric of skepticism, Davidson also has a reply to the noetic and aletheia skeptic in his argument from «On the Very Idea of a Conceptual Scheme».Foot note 7_50 The upshot of this well-known argument is that we cannot make sense of the idea of languages that are true and untranslatable into our own. If this argument is successful then the idea of other knowers like the Alpha Centurians with a transcendent conceptual scheme is incoherent.Foot note 7_51
The other line of inquiry is the role of the defeater stratagem. As noted, the argument for the viability of these other skepticisms is conditional on the acceptability of the defeater stratagem. This of course raises the question of whether the defeater stratagem is itself viable. Indeed, it seems quite possible that the line of argument canvassed here might be understood by some to constitute a reductio ad absurdum of the defeater stratagem. The idea would be to agree that the defeater stratagem forces us to move beyond justificatory skepticism to noetic and aletheia skepticism, but that these skepticisms are absurd (or perhaps simply pointless). But even if the defeater stratagem is rejected, this in itself does not show that philosophical skepticism in all its varieties is defeated. For it still must be asked whether there are other means to motivate these skepticisms that do not rely on the defeater stratagem. I have argued else, for instance, that naturalism itself (independently of the defeater stratagem) invites noetic skepticism.