The Nutter Log

Justified True Belief

By The Famous Brett Watson On Sat, 10 Apr 2004 16:07:00 +1000

The traditional formulation of propositional knowledge (in Western philosophy) involves three key components: justification, truth, and belief (JTB). Propositional knowledge is, in this tradition, a justified belief held about a truth. To elaborate, the formulation holds that three conditions are necessary, and jointly sufficient for "knowledge". First, belief: you do not know something unless you also hold it as true in your mind; if you do not believe it, then you do not know it. Second, truth: there can be no knowledge of false propositions; belief in a falsehood is delusion or misapprehension, not knowledge. Third, justification: the belief must be appropriately supported; there must be sufficient evidence for the belief. Thus, knowledge is like a three-legged stool which cannot stand when any one leg is removed. Consider lack of belief: it may be true that Alice's twin sister has just been killed in a car accident, and the police officer reporting the fact may be sufficient evidence to warrant belief, but Alice may find herself unable to accept it, and will thus fail to know it. Lack of truth also disqualifies knowledge: the pre-Copernican belief (amply justified at the time) that heavenly bodies moved around a stationary Earth is false, and is thus not knowledge, even if educated persons of the day operated under the misapprehension that it was. Lastly, lack of justification precludes knowledge: if a charlatan fortune-teller informs Alice that she will meet the man of her dreams within a month, then this proposition isn't knowledge for Alice even if she believes it and it actually happens. Knowledge must be properly grounded, and the charlatan's claim had no grounds whatsoever. This traditional formulation is not without its problems. One could argue, for example, that "knowledge", so defined, is not a very interesting concept: the individual questions of whether a proposition is true, whether a subject believes it, and whether the subject is justified in doing so do not become more interesting when the answers happen to be uniformly affirmative. Or one could argue the pragmatic case that "knowledge" is not a useful concept: it's all very well to ponder whether subject S knows proposition P given a hypothetical situation with specified truths, but what of knowledge in the real world, where determining the truth of P is part of the problem? More significantly, perhaps, one could argue that JTB is not actually an entirely sufficient account of knowledge; that situations arise in which a justified true belief is not knowledge. Edmund Gettier makes a famously disruptive case for this view in a short paper entitled, "Is Justified True Belief Knowledge?" (originally published in Analysis, 1963, pp. 121-3). Consider the following scenario from that paper. Smith

and Jones are candidates for a job, and Smith believes that (a) Jones will get the job, and (b) Jones has ten coins in his pocket. Smith's belief in both these propositions is justified: a company executive has informed him that Jones will be hired, and he's seen the coins in question. Based on these justified beliefs, Smith also believes (quite justifiably) their logical implication: (c) the person who will get the job has ten coins in his pocket. Events transpire in such a way that Jones does not get the job, despite assurances to the contrary, and the job is offered to Smith instead. As chance would have it, proposition (c) turns out to be true anyway, because Smith also had ten coins in his pocket, although he didn't realise it at the time. Thus, Smith justifiably believed proposition (c), and it turned out to be true, but did he know it? The traditional account says so, but does this still match our intuitive grasp of what knowledge entails? It seems not. One possible way of saving the JTB account from Gettier is to argue that Smith's justification for (c) was undermined, and thus he did not know (c) because his belief was not appropriately justified. Proposition (c) follows logically from (a) and (b) only if they are both true, and it turns out that (a) is false. Proposition (c) can still be true independently of both (a) and (b), as actually transpired, but Smith's grounds for belief in (c) was the truth of (a) and (b). If there is a shortfall in JTB, it is merely that we ought to have mentioned that justification must not be undermined by subsequent events. This embellishment of JTB salvages it from the given counterexample by denying the presence of justification, but other Gettier-style counterexamples may still prove problematic. More than anything else, this saving measure serves to demonstrate how much wriggle-room exists in the "justification" component, and that makes it a more intrinsically interesting concept (to my mind) than its possible by-product, "knowledge".

The Gettier Problem

By Paul Pardi
The Object of Belief What is the nature of a believing act? How is the content of a belief related to a speech act? How do beliefs refer to things in the world? In its most simple formulation, believing is having a mental state that consists of a representation of the world[1]. Many philosophers describe belief in terms of an acceptance that the world is a specific way but I think this way of speaking implies that the believer is somehow actively doing something in the act of believing. While this true in many cases, its too restrictive for my purposes. A belief is an apprehension of the world and need not include an affirmation

of it (though belief generally does have important implications for behaviora qualification well examine later). Beliefs are usually described in terms of propositions where the belief is of the proposition and a proposition is a that clause which can be written or spoken as a fact about the world. I believe that Megan had steak for dinner or you believe that Aidan is playing his guitar. The phrase after the that is a description of the world that can be expressed linguistically and this description is the object of belief. That is, the object of belief is, in an important sense, the sentence or some set of properties of the sentence and it is this that of which the proposition consists.[2] This way of understanding belief is the genesis of a whole host of mistakes that have fueled Gettier-style counterexamples to JTB and is going to be the focus of my analysis. Since speaking of belief in this way typically is established (or, in most cases, assumed) at the beginning of an analysis of knowledge, this small mistake sets things off on the wrong path at the beginning and ends up leading to much larger errors that resist correction at the end. Instead of thinking about beliefs as being about descriptions of the world, I think its more correct to say beliefs are representations of the world and propositions are the content of those representations. Beliefs are a mental state consisting of a way the world possibly could be. On this model, propositions are not the object of belief but the content. Propositions make up the content of the belief itself. In order to make this idea clearer, some philosophers like Shope prefer to talk about belief in terms of states of affairs rather than propositions where states of affairs just are possible ways the world could be. These distinctions tend to be subtle and philosophy since the linguistic turn has used the term proposition in a variety of ways that are at best inconsistent. Since propositions are generally held to be at the center of a believing act and since belief is an essential component of knowledge (and thus critical to understanding Gettier cases), we will need to spend some time analyzing belief with the goal of providing some clarity as to the nature and object of the believing act. Index

*1+ This way of speaking assumes a distinction between the mind and the world and is something I wont argue for here. This assumption is trivial however, because on a non-referential or postmodernist understanding of belief, Gettier cases carry no water to begin with. [2] See, for example, (Ferre, 1961). Ferre argues that the locus of philosophy as a discipline is the analysis of the meaning of language (p. 6). On Ferres definition, belief is completely removed from the picture and language becomes the sole object of analysis. For Ferre, words in a propositional context are the locus of meaning and the philosophers job is to unpack this meaning. Sentences, on this view, appear not to be representative but objective.

In Gettiers Wake
JOHN TURRI> john.turri@gmail.com

2. Gettier Cases and Their Structure

Chapter 10 of this volume already introduced us to Gettiers discussion. But lets expand on what was said there. Gettier presented two cases that he thought were clear counterexamples to the JTB theory. In particular, Gettier contended that his cases showed that having a justified true belief was insufficient for knowledge, from which it follows trivially that KJTB. A case of this sort is called a Gettier case. The Gettier problem is the problem of identifying why the subject in a Gettier case lacks knowledge. It is widely assumed that unless we solve the Gettier problem, well be unable to adequately define knowledge. Interestingly, Gettier wasnt the first to come up with what we now call Gettier cases. According to Bimal Matilal (1986: 1357), the classical Indian philosopher Sriharsa constructed similar examples in the 1100s to confound his opponents, and Roderick Chisholm (1989) reminds us that Bertrand Russell and Alexius Meinong also constructed such cases decades earlier than Gettier. But still it is customary to call them Gettier cases. Gettier cases are easy to construct, once you get the feel for them. Here are two prototypical examples (not Gettiers originals).
(LAMB) One of Dr. Lambs students, Linus, tells her that he owns a Lamborghini. Linus has the title in hand. Dr. Lamb saw Linus arrive on campus in the Lamborghini each day this week. Linus even gave Dr. Lamb the keys and let her take it for a drive. Dr. Lamb believes that Linus owns a Lamborghini, and as a result concludes, At least one of my students owns a Lamborghini. As it turns out, Linus doesnt own a Lamborghini. Hes borrowing it from his cousin, who happens to have the same name and birthday. Dr. Lamb has no evidence of any of this deception, though. And yet its still true that at least one of her students owns a Lamborghini: a modest young

1. Introduction
One main goal of epistemology is to define knowledge. Legend has it that the traditional or standard view of knowledge is justified true belief (K=JTB) and that this traditional view reigned supreme for decades, centuries even. As one leading epistemology textbook puts it,
It is reasonable to say [that] some version or other of the traditional conception of knowledge was taken for granted . . . by virtually all philosophers seriously concerned with knowledge in the period from the time of Descartes until the middle of the twentieth century. (BonJour 2001: 43)

But that all changed in 1963 when an unheralded young philosopher at Wayne State University in Detroit, Edmund Gettier, published a paper as short as it has been influential: Is Justified True Belief Knowledge?. Gettiers paper has since engendered a half century worth of responses. If you added up the number of times that this article has been discussed or cited in the literature (thou-sands of times), and divided that by the number of words in the article (approximately 930), the resulting quotient would be larger than the quotient for any other work of philosophy ever published. Now if we call this the citation per word formula for calculating a publications influence, then its safe to say that Gettiers article is the greatest philosophical caper of all time.

woman who sits in the back row owns one. She doesnt like to boast, though, so she doesnt call attention to the fact that she owns a Lamborghini. (Vaguely modeled after Lehrer 1965: 16970)

Most philosophers who consider this case say that (a) Dr. Lamb does not know that at least one of her students owns a Lamborghini, even though (b) she has a justified true belief that at least one of her students owns a Lamborghini.
(SHEEP) Shep is trekking through a pasture. He gazes down across the field and notices an animal. Viewing conditions are optimal. It appears to be an unremarkable sheep, so he believes, Thats a sheep in this field, from which he concludes, Theres at least one sheep in this field. As it turns out, Shep isnt looking at a sheep, but rather a dog dressed up to look just like a sheep a very cleverly disguised dog! Shep has no evidence of this deception. The thought that it was a disguised dog never even occurs to him. And yet its still true that there is a sheep in the field. Its directly behind the dog, hidden from Sheps view. (Adapted from Chisholm 1989: 93)

a justified false belief that P (which is possible, if justification isnt factive). Then have the protagonist deduce a true consequence, Q, of the justified belief that P, and have the protagonist believe Q on the basis of this deduction (surely this is possible). The resulting belief in Q will be justified (by the assumption that justification is closed under deductive entailment). And the overall result will be a justified true belief that Q, without knowledge that Q. Here is another way of understanding the Gettier recipe (Zagzebski 1996). Start with a belief sufficiently justified to meet the justification requirement for knowledge. Then add an element of bad luck that would normally prevent the justified belief from being true. Lastly add a dose of good luck that cancels out the bad, so the belief ends up true anyhow. The justification of the justified true belief appears oddly disconnected from the truth, and the overall result will be a justified true belief, one which doesnt amount to knowledge.

Most philosophers who consider this case say that (a) Shep does not know that there is a sheep in the field, even though (b) he has a justified true belief that there is a sheep in the field. Gettier claimed that the success of these cases as counterexamples to the JTB theory depends on two principles. First, that justification isnt factive. This means that its possible to have a false justified belief. Second, that justification is closed under deductive entailment. This means that if youre justified in believing some proposition P, and P entails some other proposition Q, and you deduce Q from P, and believe Q based on that deduction, then youre justified in believing Q too. In light of these two points, here is one way to understand the recipe for generating Gettier cases (Feldman 2003: 28). Begin with

3. Some Proposed Solutions to The Gettier Problem

Attempted solutions to the Gettier problem are legion. Some responses are conservative, in that they hew closely to the original JTB theory, introducing as little change as possible to handle the cases. Some responses are radical, in that they break decisively with the spirit of the JTB account, either dramatically refashioning the justification requirement, or even eliminating it entirely. Here Ill review some of the most influential and interesting responses to the Gettier problem (see Shope 1983 and Hetherington 2009 for detail on other approaches).

Some philosophers looked at Gettier cases and thought that the problem amounts to this: the subject has a justified true belief, but the belief is essentially based on a false premise. In LAMB the false premise is Dr. Lambs belief, My student Linus owns a Lamborghini, and in SHEEP it is Sheps belief, This animal in the field is a sheep (where this refers to the cleverly disguised dog). This suggests the no essential false basis theory of knowledge: (NFB) You know that P just in case (i) P is true, (ii) you believe that P, (iii) your belief that P is justified, and (iv) your belief that P isnt essentially based on any falsehood. (For examples of this idea, see Harman 1973 and Clark 1963). One problem with NFB is that it cant handle simple variants of Gettier cases. Consider this variant of LAMB.
(LUCKY LAMB): The case is the same as LAMB, except that unbeknownst to Linus he has just inherited a Lamborghini. His cousin died and left it to him.

(For examples of this line of thought, see Garfield 2005, Saunders and Champawat 1964). Other philosophers looked at Gettier cases and thought that the problem is this: the subject has a justified true belief, but the justification is defeated (see, for example, Lehrer and Paxson 1969, Klein 1976). In LAMB Dr. Lambs justification is defeated by the fact that Linus is deceiving her (or, in LUCKY LAMB, that LINUS is trying hard to deceive her). In SHEEP Sheps justification is defeated by the fact that hes being deceived by a cleverly disguised dog. This suggests the simple defeasibility theory of knowledge: (SDT) you know that P just in case (i) P is true, (ii) you believe that P, (iii) your belief that P is justified, and (iv) your justification for believing P is undefeated. Some fact F defeats your justification for believing P just in case (i) you believe P based on evidence E, (ii) E justifies belief in P, but (iii) the combination (E+F) fails to justify belief in P. In LAMB the defeater is the fact that Linus is deceiving Dr. Lamb about owning a Lamborghini (or, in LUCKY LAMB, is earnestly trying to deceive Dr. Lamb). In SHEEP the defeater is the fact that Shep is looking at a cleverly disguised dog. One problem with SDT is that it seems to rule out too much. Consider:
(INSANE) You were just tenured! Excitedly you phone to tell your best friend Sophia the wonderful news. Naturally Sophia believes and congratulates you. However, unbeknownst to either of you, your dean just went insane succumbed to the pressures of profit-driven university governance and is absolutely certain you were not tenured.

In this case Dr. Lambs belief My student Linus owns a Lamborghini is true, so NFB cant handle LUCKY LAMB, because in this case its true that Linus owns the Lamborghini in question. Another problem with NFB is that it appears to give the wrong verdict in cases like this:
(BLUE DRESS) Bill awaits Monicas arrival. He wonders whether shell wear a scarlet dress. He hears a step on the staircase and swings around to see Monica enter the room. What a dazzling indigo dress! he thinks, and concludes, Monicas dress isnt scarlet. And hes right: her dress isnt scarlet. But it isnt indigo either. Its ultramarine.

Intuitively Bill knows that Monicas dress isnt scarlet. But his belief is based on the falsehood that her dress is indigo, so NFB rules that he doesnt know that Monicas dress isnt scarlet.

It is a fact that your dean is absolutely certain that you were not tenured, and that fact combined with your testimony fails to support

Sophias belief that you were tenured, technically defeating her justification. Thus SDT rules that Sophia doesnt. But intuitively she does know. In response it has been suggested that knowledge is ultimately undefeated justified true belief. Call this the modified defeasibility theory: (MDT) you know that P just in case (i) P is true, (ii) you believe P based on evidence E, (iii) E justifies belief in P, and (iv) E is ultimately undefeated. E is ultimately undefeated just in case there is no fact F such that (E+F) fails to justify belief in P; or if there is such a fact, then there is some further fact F* such that (E+F+F*) does justify belief in P.i In such a case F* is a defeater defeater. In INSANE, F* is the fact that your deans conviction is borne of insanity. The main problem with MDT, however, is that the very device it introduces to give the intuitively correct verdict in INSANITY also deprives it of the ability to handle the original Gettier cases. Consider LAMB. The fact that Linus is deceiving Dr. Lamb is a defeater (=F). But the fact that the modest female student does own a Lamborghini is a defeater defeater (=F*). This last fact is a defeater defeater because this combination:
E: My student Linus has possession of this Lamborghini, drives it frequently, and has a title to the Lamborghini with his name and birthdate on it. F: My student Linus does not own this Lamborghini. F*:That young female student of mine owns a Lamborghini.

justifies Dr. Lambs belief that at least one of her students owns a Lamborghini. It does this because F* obviously entails that at least one of her students owns a Lamborghini. And it would do so, no matter how many of Dr. Lambs other students dont own a Lamborghini.

The responses weve looked at so far have been conservative. They respond to Gettier cases by adding a fourth condition to the three conditions featured in the traditional JTB account. But there were more radical responses. Some philosophers looked at Gettier cases and thought that the problem amounts to this: the fact that P doesnt cause the subject to believe that P (Goldman 1967). In LAMB its a fact that one of Dr. Lambs students owns a Lamborghini, but it isnt this fact (namely, the fact that the female student owns one) that causes Dr. Lamb to believe that one of her students owns a Lamborghini. In SHEEP its a fact that there is a sheep in the field, but it isnt this fact that causes Shep to believe it. This observation led to the causal theory of knowledge: (CTK) you know that P just in case (i) P is true, (ii) you believe that P, and (iii) thefact that P is true causes you to believe that P. CTK gives up on justification entirely as a condition on knowledge. The problem with CTK is that it is easy to introduce deviant causal chains into the description of any Gettier case, which would make it true that the relevant fact causes the subject to believe that P. For example, suppose that we add the following background to SHEEP. A clever farmer dressed up the cleverly disguised dog to fool Shep. The farmer was caused to do this, oddly enough, by the fact that there is at least one sheep in the field as Shep treks by. So the fact that there is at least one sheep in this field (=P) caused the farmer to dress up the dog, which caused Shep to believe that there is at least one sheep in this field. So, the fact that P caused Shep to believe that P. CTK thus rules that Shep knows that P. But intuitively this is the wrong verdict. Another problem with CTK is that if we impose a causal requirement on knowledge, then it becomes difficult to avoid skeptical

consequences for beliefs about abstract matters, such as mathematical and logical truths, because it isnt clear that, say, the fact that 2+2=4 can cause anything. It also becomes difficult to explain how we know things about the future, because it doesnt seem possible for future facts to cause our beliefs now. Strategies for overcoming these problems have been proposed, but not to the satisfaction of many. A descendant of CTK is reliabilism about justification and knowledge.ii Rather than give up entirely on justification as a condition on knowledge, some argued that we can understand justified belief as belief produced by a reliable cognitive process (Goldman 1979), and then understand knowledge as roughly justified true belief, where justification is given the relevant reliabilist reading. The result is process reliabilism: (PR) you know that P just in case (i) P is true, (ii) you believe that P, and (iii) your belief that P is produced by a reliable cognitive process (i.e. your belief is justified). PR might handle Gettier cases by pointing out that in LAMB, for example, Dr. Lambs belief is produced by making deductions based on the testimony of someone who is trying to deceive her, which plausibly isnt a reliable process. And in SHEEP Sheps belief is produced by making deductions based on mistaken appearances, which plausibly isnt reliable either. The main criticism of PR is that it has no principled way of individuating cognitive processes, and so no principled way of deciding whether any given true belief amounts to knowledge (Conee and Feldman 1998). For example, why say that Shep is basing his deductions on misleading appearances, rather than on perceptual experience? Perceptual experience is reliable. But then why doesnt

Shep know? Another problem with PR is that it cant handle simple variants of Gettier cases. Consider this variant of SHEEP.
(SPECIAL DOG) The case is the same as SHEEP, except that the disguised dog is very special. It tracks Shep and appears to him only when at least one sheep is nearby. It wouldnt appear to him unless there were a nearby sheep. It also prevents him from encountering any other non-sheep sheep look-alike that would mislead him into concluding that there is at least one sheep nearby.

In SPECIAL DOG the following method seems perfectly reliable for Shep: from the fact that something looks like a sheep nearby, conclude that there is at least one sheep nearby. So PR rules that in SPECIAL DOG Shep knows that there is at least one sheep nearby. But it would be very surprising if Shep knew in SPECIAL DOG but not in SHEEP. Other philosophers looked at Gettier cases and thought that the problem is this: its just an accident that the subjects belief is true (e.g. Unger 1968). In LAMB its just an accident that Dr. Lamb ended up being right that at least one of her students owned a Lamborghini. And in SHEEP its just an accident that Shep ended up being right that theres at least one sheep in the field. This suggests the no-accident theory of knowledge: (NAT) you know that P just in case (i) P is true, (ii) you believe that P, and (iii) it is not at all an accident that your belief that P is true. NAT omits justification from its definition of knowledge, which leaves open the possibility that there can be unreasonable knowledge, that is, knowledge which the subject is unjustified in believing is true (Unger 1968: 164). For example, if an epistemic guardian angel watched over you and ensured that your every wish came true, then wishful thinking would be a way for you to gain knowledge, since it would be no accident that your wishful beliefs turned out to be true. Many judge this to

be an absurd consequence of the view surely believing something because you want it to be true isnt a way of gaining knowledge! Another potential problem with NAT is that its very difficult to explain what clause (iii) amounts to. A related family of views propose a safety condition on knowledge (Sosa 1999, Pritchard 2005, which is intended to give content to the idea that knowledge cant be accidental or lucky. The most conservative version of a safety-based view simply appends a safety condition to the traditional analysis, yielding the safe justified true belief theory of knowledge: (SJTB) you know that P just in case (i) P is true, (ii) you believe that P, (iii) your belief that P is justified, and (iv) your belief that P is safe. A true belief is safe just in case it wouldnt easily have been false. What does it mean to say that a true belief wouldnt easily have been false? Theres no precise way to define this, but the intuitive idea is that something significant would have had to change in order to have made the belief false. One problem with this view is that it fails to handle simple variants of Gettier cases, such as SPECIAL DOG, because the Gettier-ed belief is not only justified and true, but also safe. To see why, recall that in SPECIAL DOG the cleverly disguised special dog wouldnt appear to Shep unless there were a nearby sheep, and also prevents Shep from encountering any other non-sheep sheep look-alike that would mislead him into concluding that there is at least one sheep nearby. In effect, the special dog acts as a sort of epistemic guardian angel for Shep on such matters, which ensures that his beliefs about nearby sheep are not only true but also safely formed. Another family of views proposes a sensitivity condition on knowledge (Dretske 1970, 2005, Nozick 1981). A conservative sensitivity-based view might simply append a

sensitivity condition to the JTB analysis, but sensitivity theorists typically dispense with justification altogether. Your belief that P is sensitive just in case the following conditional is true: if P were false, then you wouldnt believe that P. A sensitivity condition on knowledge handles standard Gettier cases. In LAMB if it were false that at least one of the students owned a Lamborghini, then Dr. Lamb would still believe that at least one student did (because Linus would still have deceived her). In SHEEP if it were false that there was at least one sheep in the field, then Shep would still believe that there was (because the cleverly disguised dog would still have tricked him). One problem facing this diagnosis is that it cant handle simple variants of the cases. For example, it cant handle SPECIAL DOG because the special cleverly disguised dog wouldnt have tricked Shep if there were no sheep nearby; and the special cleverly disguised dog would prevent anything else from tricking Shep; so if there werent a nearby sheep, Shep wouldnt believe that there was one. Another serious problem facing this view is that it implies that knowledge isnt closed under some trivial, known deductive entailments. (Well soon return to closure and how counterintuitive it can be to deny it, in the next paragraph and again in section 3 below. See also Vogel

1990 and Hawthorne 2005 for a defense of closure.) Related to the sensitivity-based account of knowledge, Fred Dretske (2005) has also argued for a sensitivitybased account of reasons or justification. On this view, justification isnt closed under15 | John Turri deductive entailment, and even fails to transmit across some simple, known deductive entailments. Generally speaking, a justification to believe P is an indication that P is true. Indications carry information. Information comes from sources. Consider a thermometer, which is a source of information. The thermometer indicates the ambient temperature in the room. Its readout provides a reason for believing that it is twenty-one degrees in here. The readout carries information about the ambient temperature, in this case that it is twenty-one degrees. That it is twenty-one degrees entails that it is not eighteen degrees being misrepresented as twenty-one degrees. But the latter claim is not part of the readouts content it doesnt carry that information. And yet the readouts content entails it. So your reason for believing that its twentyone degrees neednt

also be a reason for you to believe the obvious deductive consequences of the claim that its twenty-one degrees. It might be easier to grasp how potentially counterintuitive this is by considering the matter more schematically. According to Dretske, the following is possible: reason R justifies you in believing P, and you know that the truth of P guarantees the truth of Q, but still, R does not justify you in believing Q. If Dretske is right about this, then one of the key assumptions of Gettiers original discussion namely, that justification is closed under deductive entailment is thrown into doubt. 3 3 The denial of closure has significant implications for epistemology, beyond the Gettier problem. It would also enable a direct and powerful response to many influential skeptical arguments. In fact, this is precisely how the idea of denying closure entered the contemporary discussion. See Dretske 1970, Nozick 1981, and also Pritchard 2008 for a helpful overview.In Gettiers wake | 16 A more recent approach to the Gettier problem is to argue that

knowledge can be defined as true belief for which the subject earns credit for believing the truth, but a Gettier subject doesnt earn credit for believing the truth, which explains why she doesnt know (e.g. Greco 2003, Zagzebski 2009). For example, in SHEEP Shep doesnt earn credit for believing the truth about whether theres a sheep in the field. Rather, we would credit a confluence of odd circumstances for the fact that Shep ends up believing correctly. Its an open question whether the operative notion of credit can ultimately sustain this treatment of the Gettier problem. A related view defines knowledge as follows: you know that P just in case you have a true belief that P because you believed competently; however, it is argued, although the Gettier subject has a true belief, and believes competently, he doesnt have a true belief because he believes competently, which explains why he doesnt know (Sosa 2007). Another related view defines knowledge as follows: you know that P just in case the fact that you have a true belief that P is a manifestation of your cognitive powers; however, it is argued, although the Gettier

subject has a true belief, and exercises her cognitive powers, the fact that she has a true belief isnt a manifestation of her cognitive powers, which explains why she doesnt know (Turri 2011, forthcoming). The jury is still out on this family of approaches. Some have argued that Gettiers intuition about his cases was wrong: Gettier cases are cases of knowledge. Stephen Hetherington (1998, 1999, 2011) argues that a Gettier subject knows despite coming perilously close to not knowing, and supplements this by dia-17 | John Turri gnosing intuitions to the contrary. Whereas safety theorists would claim that the unsafety of a Gettier subjects belief disqualifies it as a case of knowledge, Hetherington contends that its unsafety misleads us into thinking that the Gettier subject doesnt know. The Gettier subjects belief might very easily have been false, and we mistake this near failure for an actual failure. Gettier subjects straddle the divide between just barely knowing and not knowing. Although ingenious, Hetheringtons view remains a minority position (Lycan 2006, Turri forthcoming).

Many philosophers have taken Gettier cases to show that justified true belief isnt sufficient for knowledge, even though it still is necessary. And as weve already seen, some have tried to replace justification with something else entirely, such as an appropriate casual relation, a safety condition, or a sensitivity condition. But at least one philosopher has argued that theyre all wrong because knowledge is simpler than any of them had imagined: knowledge is mere true belief (Sartwell 1991, 1992). Crispin Sartwells argument for this position is simple: knowledge is the goal of inquiry; the goal of inquiry is true belief; so knowledge is true belief. Inquiry just is the procedure of generating beliefs about particular propositions, and when we ask whether some claim is true, what we want is to know whether its true. In other words, knowledge is the goal of inquiry. But most philosophers will object that we also want our true beliefs to be justified, well supported by evidence, so Sartwell has left out an important aspect of our goal. Sartwell accepts that we want justified beliefs, but argues that this is only because justifica-In Gettiers wake | 18

tion is a good sign that weve got what we really want, namely, true belief. Justification is instrumentally good because its a good sign that we do know, but isnt an essential part of knowledge. 4. The Scylla and Charybdis of post-Gettier epistemology: Or, teetering between fallibilism an skepticism Nearly all epistemologists think that Sartwell is wrong, and that knowledge requires something more than true belief. Most epistemologists still think that justification is a necessary condition on knowledge, even if justified true belief isnt sufficient for knowledge. And most epistemologists still agree with Gettier that justification isnt factive (Sutton 2007 dissents). Having a justified belief that P doesnt guarantee that P is true: you could be justified in believing P even though P is false. Moreover, it is widely held that the minimum level of justification required for knowledge is also non-factive: having knowledge-grade justification for believing that P doesnt guarantee that P is true. To put it differently, the conventional wisdom in contemporary epistemology is that knowledge-grade justification is fallible: you could be wrong even though you have it.

But fallibilism has struck many as deeply problematic. What follows is one way of explaining why fallibilism can seem both attractive and deeply puzzling (BonJour 2001). Suppose you have a true belief. In order for it to be knowledge, how much justification must be added to it? Think of justification for a belief as measured by how probable the belief is given the19 | John Turri reasons or evidence you have. We can measure probability any way we like, but one convenient way to measure it is to use the decimals in the interval [0, 1] on the number line. A probability of 0 means that the claim is guaranteed to be false. A probability of 1 means that the claim is guaranteed to be true. A probability of .5 means that the claim is just as likely to be true as it is to be false. The question then becomes: how probable, relative to your evidence, must your belief be for it to be knowledge? Obviously it must be greater than .5 after all, if it were less than .5, then it would be more probable that your belief was false, given your evidence! But how much greater than .5? Suppose we say

that knowledge requires a probability of 1 that is, knowledge requires justification that guarantees the truth of the belief. Call this infallible justification. The infallibilist conception of knowledge says that knowledge requires infallible justification. We can motivate the infallibilist conception as follows. If the aim of belief is truth, then it makes sense that knowledge would require infallible justification, because it guarantees that beliefs aim is achieved. Clearly its a good thing to have such a guarantee. But all is not well with the infallibilist conception. It seems to entail that we know nothing at all about the material world outside of our own minds, or about the (contingent) future, or about the (contingent) past. For it seems that we could have had the same justification that we do in fact have, even if the world around us (or the past, or the future) had been radically different. Our justification doesnt guarantee that a material world exists. (Think of DescartessIn Gettiers wake | 20 evil genius.) Neither does it guarantee that there is a past or future. This dramatic skeptical consequence conflicts with commonsense and counts against the infallibilist conception of knowledge. This is

presumably part of the motivation for the widespread agreement that justification isnt factive. We seem compelled to conclude that knowledge requires justification that makes the belief very likely true, but neednt guarantee it. This is the fallibilist conception of knowledge. But a question about this view immediately arises: what level of justification does it require? Any point short of 1 would seem arbitrary. Why should we pick that point exactly? The same could be said for a vague range that includes points short of 1 why, exactly, should the vague range extend that far but not further? This might not seem so troubling in itself, but as Laurence BonJour (2001) points out, it suggests an even deeper problem for the weak conception. It brings into doubt the value of knowledge. Can knowledge really be valuable if it is arbitrarily defined? It would count heavily against the fallibilist conception of knowledge if it implied that knowledge wasnt valuable. (Kaplan 1985 raises related worries about knowledges value in light of the Gettier problem.) A related problem for the fallibilist conception of knowledge

presents itself, which relates to the second of Gettiers assumptions. Suppose for the sake of argument that we settle on .9 as the required level of probability. Suppose further that you believe Q and you believe R, that Q and R are both true, and that you have reached the .9 threshold for each. Thus the fallibilist conception en-21 | John Turri tails that you know Q, and it entails that you know R. Intuitively, if you know Q and you also know R, then you know the conjunction (Q&R), just by simple deduction. But, surprisingly, the weak conception of knowledge cant sustain this judgment! To see why, consider that the probability of the conjunction of two independent claims, such as Q and R, equals the product of their probabilities. (This is the special conjunction rule from probability theory.) In this case, the probability of Q = .9 and the probability of R = .9. So the probability of the conjunction (Q&R) = .9 .9 = .81, which falls short of the required .9. So the weak conception of knowledge along with a law of probability entail that you dont know the conjunction (Q&R), because you arent well enough justified in believing the conjunction. Can we tolerate this result?

So we are faced with a choice between two views, fallibilism and infallibilism, each of which has seemingly unpalatable consequences. If we accept fallibilism, then we seem poised to surrender the intuitive claim that (knowledge-grade) justification is closed under simple, known deductive entailments, and also the intuitive claim that knowledge is valuable. And if we accept infallibilism, then we seem poised to surrender the intuitive claim that were in a position to know lots of things about the material world, the past and the future. Notice how Gettiers two assumptions relate to these unpalatable consequences. In setting up his problem, Gettier assumed that (1) justification isnt factive, and (2) justification is closed under deductive entailment. Infallibilism threatens to falsify something inIn Gettiers wake | 22 the ballpark of (1), whereas fallibilism threatens to falsify something in the ballpark of (2). Gettiers lasting legacy might well be to force us to choose between these two claims. Are we forced to choose between them, or can we find some way to have our epistemological cake and eat it too? 4 4

For helpful feedback, I thank Stephen Hetherington and Angelo Turri.23 | John Turri References BonJour, Laurence. 2001. Epistemology: Classical Problems and Contemporary Responses. Lanham, Maryland: Rowman and Littlefield. Chisholm, Roderick M. 1989. Theory of Knowledge, 3 rd ed. Englewood Cliffs, NJ: Prentice Hall. Clark, Michael. 1963. Knowledge and Grounds: A Comment on Mr. Gettiers Paper. Analysis 24.2: 4648. Conee, Earl and Richard Feldman. 1998. The Generality Problem for Reliabilism. Philosophical Studies 89.1: 129. DePaul, Michael and Linda Zagzebski, eds. 2003. Intellectual Virtue: Perspectives from Ethics and Epistemology. Oxford: Oxford University Press. Dretske, Fred. 1970. Epistemic Operators. The Journal of Philosophy 67.24: 10071023. Dretske, Fred. 2005. The Case Against Closure. In Contemporary Debates in Epistemology. Ed. Matthias Steup and Ernest Sosa. Malden, Mass.: Blackwell.

Feldman, Richard. 2003. Epistemology. Upper Saddle River, NJ: Prentice Hall. Gettier, Edmund. 1963. Is Justified True Belief Knowledge? Analysis 23.6: 1213. Goldman, Alvin I. 1979. What Is Justified Belief? In Justification and Knowledge. Ed. George Pappas. Dordrecht: Reidel. Greco, John. 2003. Knowledge as Credit for True Belief. In DePaul and Zagzebski, eds. Harman, Gilbert. 1973. Thought. Princeton: Princeton University Press. Hawthorne, John. 2005. The Case for Closure. In Contemporary Debates in Epistemology. Ed. Matthias Steup and Ernest Sosa. Malden, Mass.: Blackwell. Hetherington, Stephen. 1998. Actually Knowing. The Philosophical Quarterly 48.193: 453469.In Gettiers wake | 24 Hetherington, Stephen. 1999. Knowing Failably. The Journal of Philosophy 96.11: 565587. Hetherington, Stephen, ed. 2006. Epistemology Futures. Oxford: Oxford University Press. Hetherington, Stephen. 2009. The Gettier Problem. In The Routledge Companion to Epistemology. Ed. Sven Bernecker and Duncan Pritchard. New York: Routledge.

Hetherington, Stephen. 2011. How to Know: A Practicalist Conception of Knowledge. Malden, Mass.: Wiley-Blackwell. Kaplan, Mark. 1985. Its Not What You Know that Counts. The Journal of Philosophy 82.7: 35063. Klein, Peter. 1976. Knowledge, Causality, and Defeasibility. The Journal of Philosophy 73.20: 792812. Lehrer, Keith. 1965. Knowledge, Truth and Evidence. Analysis 25.5: 168175. Lehrer, Keith and Thomas Paxson Jr. 1969. Knowledge: Undefeated Justified True Belief. Journal of Philosophy 66: 225237. Lycan, William G. 2006. On the Gettier Problem problem. In Epistemology Futures. Ed. Stephen Hetherington. Oxford: Oxford University Press. Matilal, Bimal Krishna. 1986. Perception: An Essay on Classical Indian Theories of Knowledge. Oxford: Oxford University Press. Nozick, Robert. 1981. Philosophical Explorations. Cambridge, Mass.: Harvard University Press. Pritchard, Duncan. 2005. Epistemic Luck. Oxford: Oxford University Press.

Pritchard, Duncan. 2008. Sensitivity, Safety, and Anti-Luck Epistemology. In The Oxford Handbook of Skepticism. Ed. John Greco. Oxford: Oxford University Press. Sartwell, Crispin. 1991. Knowledge Is Merely True Belief. American Philosophical Quarterly 28.2: 15765. Sartwell, Crispin. 1992. Why Knowledge Is Merely True Belief. The Journal of Philosophy 89.4: 167180.25 | John Turri Saunders, John and Narayan Champawat. 1964. Analysis 25.1: 89. Shope, Robert. 1983. The Analysis of Knowing: A Decade of Research. Princeton, NJ: Princeton University Press. Sosa, Ernest. 1999. How to Defeat Opposition to Moore. Philosophical Perspectives 13: 141 53. Sosa, Ernest. 2007. A Virtue Epistemology: Apt Belief and Reflective Knowledge, v. 1. Oxford: Oxford University Press. Sutton, Jonathan. 2007. Without Justification. Cambridge, Mass.: MIT Press.

Unger, Peter. 1968. An Analysis of Factual Knowledge. The Journal of Philosophy 65.6: 15770. Vogel, Jonathan. 1990. Are There Counterexamples to the Closure Principle? In Doubting. Ed. M.D. Roth and G. Ross. Dordrecht: Kluwer. Warfield, Ted. A. 2005. Knowledge from Falsehood. Philosophical Perspectives 19: 405416. Zagzebski, Linda. 1996. Virtues of the Mind: An Inquiry into the Nature of Virtue and the Ethical Foundations of Knowledge. Cambridge: Cambridge University Press. Zagzebski, Linda. 2009. On Epistemology. Belmont, Calif.: Wadsworth.
i

This is an oversimplification, since there might be defeaters for a defeater defeater. But still, the point is clear enough: whenever justification is defeated (or a defeater defeater is defeated) theres always at least one otherfact to defeat the defeater (or to defeat the defeater defeater).
ii

Turri, John. 2011. Manifest Failure: The Gettier Problem Solved. Philosophers Imprint 11.8: 111. Turri, John. Forthcoming. Is Knowledge Justified True Belief? Synthese.

Both reliabilism and the CTK were discussed also in Chapter 10 of this volume.