Epistemology

If TAK were correct, conditions (i)–(iii) would be sufficient for knowledge. Using a couple of clever examples, Edmund Gettier showed that they are not. Ever since, it has become common practice to refer to cases that demonstrate the insufficiency of (i)–(iii) as "Gettier cases." Here is a simple example. Suppose S believes that there is a sheep in the field. But what S takes to be a sheep is merely a rock that, bizarrely, looks exactly like a sheep. Since S has no reason to assume that he is misled in that way, he is justified in believing that there is a sheep in the field. Thus far, S's belief is justified and false. But suppose further that, behind a bush that blocks S's view, there really happens to be a sheep in the field. Given this further assumption, S's belief turns out to be a justified true belief that clearly fails to be an instance of knowledge. What condition must be added to (i), (ii), and (iii) to rule out Gettier cases? Even today, about three decades after the publication of Gettier's 1963 article, this puzzle, commonly referred to as the "Gettier problem," remains unresolved.

According to philosophers who favor traditional epistemology, the existence of the Gettier problem does not establish that TAK is without merit. It merely shows that TAK is an approximation that needs further refinement. From this point of view, if S is to know that p, S must indeed have justification, or a good reason, for believing that p. But if the justification requirement is understood as a demand for the possession of good reasons, then the following problem—known as the regress problem—arises. Suppose S's reason for believing p is q. Now, if q is to justify S's belief that p, S must have justification for accepting that q. Given that justification requires the possession of a good reason, it follows that S must have a further reason, r, for q, and that S must have a still further reason, s, for r, and so forth. Regarding the structure of this regress, there are the following possibilities: (i) it continues ad infinitum; (ii) it terminates in basic belief: a belief that is justified without receiving its justification from any other beliefs; (iii) it circles back to its origin, the belief that p. These possibilities allow for the following positions, characterized in rough outline: Skeptics endorse (i) and conclude that justification is impossible. Infinitists agree with skeptics that the regress cannot be stopped, but hold that a chain of reasons can justify even it is infinite. Foundationalists advocate (ii), and coherentists favor (iii). Neither infinitism nor skepticism enjoys broad appeal. Foundationalism and coherentism, therefore, may be considered the main contenders.

Foundationalists claim the following: (i) in addition to mediately justified beliefs—beliefs that receive their justification from other beliefs—there are immediately justified, or basic, beliefs: beliefs that are somehow justified without depending for support on any other beliefs; (ii) ultimately, all mediately justified beliefs have their justification conferred upon them by immediately justified beliefs. Accordingly, foundationalists face two main challenges. They must give a plausible and detailed account of how it is possible for a belief to be justified without receiving its justification from any other beliefs, and they need to explain precisely how basic beliefs provide justification for mediately justified beliefs. According to classical foundationalism, basic beliefs are infallible and entail the mediately justified beliefs that are inferred from them. According to more recent, modest versions of foundationalism, basic beliefs need not enjoy any epistemic privilege as strong as infallibility, and can support nonbasic beliefs without entailing them.

Coherentists claim that there are no basic beliefs. All justified beliefs receive their justification from other beliefs. The chief idea coherentists invoke is that justification is holistic. For a given belief, B, to be justified, the subject must have justification for an entire set of other beliefs that, together with B, form a coherent whole. Coherentist theories are typically defended by highlighting the difficulties involved in the idea of immediate justification. For example, Sellars points out that, when it comes to explaining how basic beliefs are justified, foundationalists face the dilemma of having to conceive of a basic belief's justification as being either propositional or nonpropositional. In the former case, the regress problem returns, for if what justifies an allegedly basic belief that p is a proposition q, the question immediately arises of what reason there is for believing that q is true. In the latter case, the problem is how we are to conceive of nonpropositional justification. Presumably, such justification arises from what is "given" in perceptual experience: some kind of non-propositional content. But, according to Sellars, if what is experientially given is nonpropositional, it simply is not the kind of thing that has the capacity to justify anything. Coherentism, then, derives its main defense from the apparent impossibility of immediate justification. Unfortunately, the alternative coherentists offer is no less problematic. Versions of coherentism that are developed in detail either suffer from circularity or regress problems, or else threaten to collapse back into foundationalism. The regress problem may, therefore, be viewed as one of the persistent and seemingly unsolvable puzzles of epistemology.

A straightforward, though certainly not uncontroversial, response to the regress problem becomes available when the traditional conception is rejected on behalf of a radical epistemological reorientation, according to which justification is a matter of, not possessing evidence or good reasons, but instead originating in causes of the right kind. Obviously, if justification does not require reasons, placing a justification condition on knowledge will not generate a regress of reasons. The key idea of the reoriented approach is that beliefs are caused by cognitive processes that are either reliable or unreliable. Perception, memory, and introspection are reliable processes. Biased thinking, wishful thinking, and making hunches are unreliable processes. According to reliabilism, this distinction is crucial for developing a successful theory of justification. The basic idea is roughly that a belief is justified (or, according to other versions, an instance of knowledge) if and only if it is caused by a reliable cognitive process—that is, a process that would produce mostly true beliefs in a wide range of different circumstances. Related theories assert, roughly, that S's belief that p qualifies as knowledge if S would not believe that p if p were false, or that S's true belief that p, on the basis of a reason r, qualifies as knowledge if S would not have r if r were false.

Advocates of TAK reject such theories on the following grounds. First, for a belief to be justified, it need not have its origin in a reliable cognitive process. Consider the victim of an evil demon. Such a subject has a belief system and evidence analogous to that of a normal person. But unlike the beliefs of a normal person, those of an evil demon victim are massively false. Reliabilism implies, implausibly, that the victim's beliefs are unjustified. Second, origination in reliable cognitive processes is not sufficient for making a belief justified, or, if true, an instance of knowledge. Advocates of TAK would argue that, for S to be justified in believing that p, or to know that p, it is necessary for S to have a good reason for p. (For a locus classicus of that kind of criticism, see BonJour, chapter 3.)

Theories that favor the traditional conception of knowledge are typically labeled "internalist." In contrast, causal theories that make de facto reliability the key notion of epistemic assessment are referred to as "externalist." Traditional theories, which identify justification with the possession of a good reason, are claimed to be internalist on the following ground: whether a subject, S, has a good reason for believing that p is something that S can determine merely by reflecting on the matter. Advocates of reliabilism and related theories are considered externalists because they deny that what determines a belief's justificational status must be something that is recognizable upon reflection.

In recent years, an approach closely related to causal theories and reliabilism—virtue epistemology—has received much attention (Sosa; Zagzebski). Virtue epistemologists advocate that justification and knowledge must be understood as arising from the employment of virtuous—that is, reliably working—cognitive faculties. Plantinga developed in detail a widely discussed theory of this kind. Its basic claim is that warrant—the property that a true belief into knowledge—must be identified with the proper functioning of faculties that are well designed, either by God or natural selection.

A central concern of contemporary epistemology, connecting back to such seminal figures as René Descartes, David Hume, Thomas Reid, George Edward Moore, and Bertrand Russell, is skepticism. The modern version of Descartes's evil demon is the mad scientist who controls the thoughts and beliefs of a brain in a vat (BIV). A BIV's evidence is, ex hypothesi, the same as yours or mine. For example, you have experiences of your hands—you see and feel them—on the basis of which you claim to know that you have hands. A BIV has hand-like experiences just as you do, but of course is mistaken in her belief that she has hands. On what grounds, then, can you claim to know that you have hands? The skeptical problem arises because the skeptical hypothesis in question—I am BIV—is incompatible with many of the ordinary propositions you take yourself to know. For example, if you are a BIV, then your current beliefs about your location, and the objects in your immediate environment, are all false. A skeptic, then, could argue as follows: You don't know that (say) you have hands unless you know that the BIV hypothesis is false. But you don't know that that hypothesis is false. Consequently, you don't know that you have hands.

In response to such skepticism, the following responses have been advocated: (i) evidentialism: I actually have evidence for believing that I am not a BIV; (ii) the relevant alternatives theory: the skeptical hypothesis fails to undermine ordinary knowledge claims since it is not a relevant alternative; (iii) inference to the best explanation: the skeptical hypothesis fails to defeat ordinary knowledge claims because it explains our beliefs and experiences less well than the hypothesis that the world is pretty much what we take it to be. A relatively new and influential response to skepticism is (iv) contextualism, according to which knowledge attributions are true or false depending on the attributor's standards of knowledge. When skeptical hypotheses are entertained, the standards of knowledge rise and become extremely stringent. As a result, a subject in such a context would be incorrect in saying she knows that she is not BIV, or that she has hands. However, in contexts in which skeptical hypotheses are ignored, the standards of knowledge remain low. It will then be correct to say that one has knowledge of one's hands. According to contextualists, what recommends this approach is that it preserves the truth of our ordinary knowledge claims and, at the same time, gives skepticism its due. For the gist of the contextualist solution is that our typical knowledge attributions, such as "I know I have hands," are in ordinary situations correct, and incorrect only in those contexts in which we concern ourselves with skeptical hypotheses. (For a collection of important essays on the problem of skepticism, see De Rose and Warfield. For a collection of essays debating a wide range of contemporary epistemological issues of the kind mentioned above, see Sosa and Steup.)