AMERICAN PHILOSOPHICAL QUARTERLY

I. INTRODUCTION

In recent years, epistemologists have become increasingly impressed with reliabilist theories of justification.¹ Reliabilism is often formulated as the claim that a belief is justified²just in case it is a reliable belief; however, this formulation can be somewhat misleading. There is a sense in which a set of beliefs can be reliable, just as a certain history or testimony can be reliable: what one means is that a certain set of propositions is highly accurate, has mostly true members, or is not wrong in any important way. Reliabilists, though, do not just want to say that a belief is justified just in case it is a member of a type with mostly true members, i.e., just in case it is probably true; they also want to appeal to the notion of reliability in that sense in which we say that persons, processes, procedures, tests, and experiments are reliable. Reliabilism is a view both about the reliability of beliefs and about the reliability of the person who has the belief or the procedure that is responsible for the belief.

So, we can distinguish two quite different central features of reliabilism. The first is its connection with truth; according to reliabilists, in order for a belief to be justified it must be objectively likely to be true. Second, reliabilists claim that what makes a belief probably true is the dependability either of the person holding the belief or of the process or procedure by which the belief either comes to be held or is (causally) sustained. Thus, we can distinguish personal reliabilism, with its emphasis on the intellectual character of the person in question with respect to accepting the truth and avoiding error, from procedural reliabilism, which emphasizes the efficaciousness of that procedure with regard to generating truth and avoiding error.

The notion of a process or procedure to which procedural reliabilists appeal is often a historical notion: it involves some sort of temporal duration.We need not require temporal duration for a version of procedural reliabilism; instead, I shall merely require that any theory involving a procedural requirement not preclude temporal duration. Hence, I shall construe the procedural requirement in a quite weak sense: only a theory which insists that it is impossible that the past affect justification denies the procedural requirement. So, for example, coherence theories normally fail to include a procedural requirement since it is normally present coherence that matters; causal theories, even those emphasizing sustaining causes, include procedural requirements since causal relations normally involve some time lapse (however minute).

Armstrong’s theory provides a good example of personal reliabilism. Armstrong claims that a person knows a truth when and only when there is a law-like connection between the truth of the belief and the person’s holding that belief.³He calls this view a “thermometer” view of knowledge. Some thermometers happen to register the correct temperature on a given occasion, yet are defective. Such thermometers are to be likened to instances of true belief which are not knowledge. Other thermometers are not defective so that a certain reading on the thermometer insures that the temperature is as the thermometer indicates. A person who knows is like a reliable thermometer: his believing a proposition in a given circumstance insures that the proposition in question is true.⁴So, for Armstrong, it is the person that is the locus of reliability.

Goldman provides an excellent example of procedural reliabilism. Goldman claims that a belief is justified if and only if it is produced by a reliable method—a method which produces mostly true beliefs.⁵On Goldman’s view, the personal reliability of the individual is not important; a person may be epistemically skewed so that it is only a random occurrence that he uses a reliable method on any given occasion. Such a person is clearly unreliable, yet the method he employs may itself be reliable.

In this paper, I wish to discuss how to capture the two sorts of intuitions that lie behind reliabilist theories of justification: the truth motivation (that a justified belief is one which is most likely true), and the dependability motivation (that either the person or the procedure used must be dependable in signalling truth). I shall propose how to capture these intuitions, for I think they are important ones; nonetheless, the way that these intuitions can be accommodated does not generate a reliabilist theory of justification. First, I shall suggest that the truth motivation can be captured only in a derivative manner from a more fundamental theory of justification. Further, even though the dependability motivation is an important insight into the cognitive ideal, no such dependability requirement should be taken to be a necessary condition for justification. Hence, I shall defend that having justified beliefs is not the intellectual ideal, though it clearly is a part of it. Let us begin with the truth motivation.

II. JUSTIFICATION AND TRUTH

The truth motivation, to repeat, claims that there must be some sort of objective connection between the justification of a belief and the truth of that belief. There is at least a grain of truth in these claims. In order to distinguish between that sort of justification which is a merely practical justification and that sort of justification—call it epistemic justification—which passes one of the hurdles on the path from belief to knowledge, it would seem that we must have reference to some connection between justification and truth which does not obtain solely in virtue of the beliefs being held. One cannot say, though, that an epistemicjustification is one that best achieves the goal of believing the truth and avoiding error, for the best way to achieve that goal is simply to believe the truth and not believe what is false. The consequence of this statement is that there is no distinction between justification and truth, hence a bit more care is called for.

Suppose we say, instead, that an epistemic justification obtains when holding the belief in question achieves the goal of finding the truth and avoiding error in virtue of some property that belief has other than being true. Such a view is better than the earlier one, in that it does not simply collapse the distinction between justification and truth. However, it is stronger than is needed to capture the truth motivation for reliabilism, for it would seem sufficient to capture that motivation that it is objectively likely that holding the belief in question achieves the goal in question. Thus, we might try:

J1: Necessarily, S’s belief that p is epistemically justified iff it is objectively likely that believing p best satisfies the goal of believing the truth and avoiding error (in virtue of having some property other than being true).

Call J1 an objectivist account of epistemic justification, and contrast it with the following subjectivist account:

J2: Necessarily, S’s belief that p is epistemically justified iff S believes that believing p best satisfies the goal of believing the truth and avoiding error.

J2 implies that there is more to the connection between justification and truth than just what obtains in virtue of S’s believing that p, for S can believe that p and yet not believe that holding this belief best achieves his goal of finding the truth and avoiding error. On the other hand, J2 obviously would not satisfy the truth motivation for reliabilism. For the sort of connection that can obtain on J2 between justification and truth does not require that anyone’s justified beliefs be true. J2 can be satisfied even if all, or mostly all, beliefs are in fact false.

Consider another account of the connection in question:

J3: Necessarily, S’s belief that p is epistemically justified if the property which makes S’s belief that p justified also makes the following proposition justified for S: believing p best satisfies the goal of finding the truth and avoiding error.

To see how J3 works, consider an example of a pragmatically justified belief: believing that one will succeed in one’s endeavors on the grounds that if one does not believe this claim, the lack of that belief is causally sufficient to insure failure. In order for this justification to be epistemic justification, these same grounds must justify the proposition that believing that one will succeed best satisfies the goal of believing the truth and avoiding error. But surely these grounds do not justify this latter claim: they don’t give any sort of reason whatsoever for thinking that this latter claim is true or false.

I shall call this version of the connection between justification and truth an iterative account of epistemic justification. The defining characteristic of an iterative account is that it denies that epistemic justification can be clarified solely by concepts other than justification itself. Thus, any clarification will have a recurrence of the same concept of justification both in the definiendum and in the definiens.

Those that affirm the truth motivation will not be any more satisfied with the iterative account than with the subjective account. For the iterative account does not establish any sort of objective link between justification and truth, any more than does the subjective account. So, reliabilists cannot be happy with J3 either.

One other possible approach to distinguishing epistemic justification from other sorts of justification is to take epistemic justification as primitive and define the other sorts of justification in terms of this primitive notion. For example, one might be able to define practical justification as that sort of justification which obtains when and only when one is epistemically justified that failing to do an action will frustrate one of one’s practical goals.⁶ But, even if such an account were to succeed, it would not satisfy the truth motivation for reliabilism. For that motivation requires that an epistemically justified belief be at least objectively likely to be true, and this account does not imply that this claim is true.

So, if the truth motivation is to be affirmed, something like J1 must be acceptable. J1 is not acceptable, though; for it is possible that there are means that get us to the truth and allow us to avoid error which we reasonably reject. For example, it might be that wishful thinking is a procedure which always gets us to the truth, even though we quite rationally believe that it does not. On 11, those that engaged in wishful thinking would thereby acquire justified beliefs, were wishful thinking to get us to the truth. However, beliefs arrived at by wishful thinking are prime examples of unjustified beliefs, hence we must reject J1.

There is a way, though, to avoid the wishful thinking case and yet maintain some objective connection between justification and truth. This can be accomplished by combining an objective account with some other account. The general strategy is to accept:

J1a: Necessarily, S’s belief that p is epistemically justified iff p has some property in virtue of which it is justified, and most beliefs which have that property are true.

J1a is an objective account of the connection between justification and truth, since it claims that a belief is justified if and only if it is most likely true. It also appears to be a version of an iterative theory, since it appeals to the notion of justification both in the definiendum and the definiens. This appearance can be misleading though, since a theory could include 11 a in addition to giving either a subjective account of justification or taking justification to be an undefined primitive. In any case, the definitive feature of J1a is that it attempts to maintain the objective account by combining it with certain features of other accounts. Further, it differs in how it handles the wishful thinking case above, for since being the product of wishful thinking is not a property which justifies beliefs, even if wishful thinking does generate true beliefs, that is irrelevant on Jla. According to J1a, only properties which both justify beliefs and signal the truth are relevant.

However, Jla is not adequate either. The best way to see this is to take some instance of a justification-conferring property—for example, consider beliefs arrived at through perceptual processes. Such beliefs are ordinarily justified and are also ordinarily true. Yet, such beliefs need not be true to be justified; conditions can be abnormal (either external, environmental conditions or internal problems with the perceptual mechanism) so that what appears to be true is not what is true, and yet the rationality of beliefs arrived at by perception not be undermined. This fact provides the clue to the inadequacy of J1a. For suppose that, in the future, some highly skilled brain surgeon, in the process of removing a brain tumor, alters the physical constituents of a person’s perceptual apparatus thereby causing the person in general to have false perceptual beliefs. Such perceptual beliefs can still be epistemically justified; nonetheless, those beliefs, though justified, are not at all likely to be true.

One might think that my objection to J1a rests on a too narrow specification of the property in virtue of which S’s belief is justified. Instead of specifying the property as the property of being produced by S’s perceptual apparatus, why not specify it as the property of being produced by a perceptual apparatus?

Well, first, it is S’s perceptual apparatus that produces the belief, not just any apparatus. But even if we accept this alteration, it will not save J1a, for it is possible to generalize the case in such a way that most perceptual apparati mislead. Perhaps a nuclear war will occur in which very few will survive, and almost all survivors need some sort of eye surgery. It might also be that the surgeon who operates on the survivors alters their perceptual apparatus as described above so that most all perceptual beliefs are false.

The last hope for defending J1a is to point out that if one counts all the perceptual beliefs over the entire history of the world, this attempt to generalize the surgeon case may not imply that most perceptual beliefs over the entire history of the world have been false. Yet, even this last attempt will not save J1a, for the example can be generalized so that both the past and future are constituted by mainly false perceptual beliefs. We are not without paradigms here, for evil demon hypotheses and brain in the vat hypotheses abound. It certainly seems possible that all of our appearance states are identical to what they now are, that all of our beliefs are just as rational or justified as they now are, and yet our beliefs be in general false because of the activity of an evil demon or the fact that we are really just brains in the vat.

And thus we end up back with Descartes. The lesson here, I think, is that evil demon worries are not limited to epistemologists with infallibilist leanings. Even if we grant that it is irrelevant whether it is possible that we are wrong about our beliefs, it does not follow that evil demon hypotheses are irrelevant. For it is in virtue of granting that the possibility of error does not undermine justification that this new evil demon problem arises. For, once we grant that the possibility of an evil demon does not matter to the justificatory status of belief, we ought also to grant the obvious truth that even if we were in a world run by an evil demon, our beliefs could be roughly as rational as they now are. Thus, to deny that the possibility of error undermines justification does not entail that evil demon worries are over: to anyone who accepts any objective version of the connection between justification and truth, the same worries will arise all over again.

Consider an example of a reliabilist aware of the problem evil demon hypotheses raise. Goldman says,

“Suppose that wishful thinking turns out to be reliable in the actual world!.. Let me.. try to give a better rendering of our aim and the theory that tries to achieve that aim. What we really want is an explanation of why we count, or would count, certain beliefs as justified and others as unjustified. Such an explanation must refer to our beliefs about reliability, not to the actual facts. The reason we count beliefs as justified is that they are formed by what we believe to be reliable belief-forming processes.. What matters, then, is what we believe about wishful thinking, not what is true (in the long run) about wishful thinking.”⁷

Goldman only explicitly considers the wishful thinking case, but the wishful thinking example is only a taste of things to come: it merely raises one sort of case along the lines of evil demon hypotheses. In addition to wondering what to say about other possible worlds where wishful thinking generates mostly truths, there is the worry that our world is one in which wishful thinking is reliable. And, in addition to the infallibilist’s worries about other possible worlds run by evil demons, there is the worry which any attempted preservation of the truth motivation must face: what if our world is the possible world the infallibilists wondered about?

We should note that Goldman’s response does not address the problem, rather it only changes the subject. Instead of talking about justification and what is necessary for it, Goldman changes the topic to a discussion of what we count as a justified belief. How seriously he means to take this paragraph is not clear, for he ignores it in the rest of the article.⁸In any case, it is clear that Goldman does not solve the evil demon problem here; he recognizes that it is a problem and just choose to change the subject instead.

What I am suggesting is that the use of evil demon considerations is a helpful way to see that the truth motivation for reliabilism is problematic. Yet, it is important to note here that I am not claiming that there is no connection between justification and truth, nor even that there is not a necessary connection between justification and truth; for we can take a lesson here from Descartes. Descartes’ solution to evil demon worries is to argue that it is logically impossible that there be such a world—it is logically impossible that there be such a world since God exists, and a good God could not allow any error to occur that we are not responsible for ourselves. Now, this latter claim is probably too strong—I can see no reason to require that we have the capacity to avoid error completely. Yet, the first part of the claim does seem right. A perfectly good Being capable of creating beings that can actually find the truth and avoid error could not remain faultless were He to create beings that in general only found error. To be perfectly good requires, first, being non-deceptive; but it also requires disclosing oneself and what one knows. Now, this requirement is not that God tell us everything He knows—perhaps in order to further our development, He just gives us the means to find the truth. The search may be hard, but we, by struggling, do find it; and we improve our intellectual characters in the process.

The strategy of such a solution to evil demon worries is to formulate an instance of the following argument: if a certain proposition p is true, then we are not in an evil demon world; p is true; hence we are not in an evil demon world. Obviously, the critical premise here is the claim that p is true; it must be is reasonable to accept that premise, i.e., there must also be an acceptable theory of justification which implies that necessarily, in certain conditions, believing p is reasonable. The catch for the objective account of the connection between justification and truth is that the justification provided for believing p by those conditions must be independent of the existence or non-existence of an evil demon in order to provide a philosophically adequate defense of the objectivity of a theory of justification. For if the conditions under which believing p is reasonable are dependent on whether there is an evil demon or not, we will not have an adequate defense of objectivity unless we can reasonably believe, independent of the existence of an evil demon, that the conditions which make believing p reasonable obtain. (This is not to say, of course, that believing p is not reasonable unless we can defend that the conditions in question are independent of the existence of an evil demon—I am talking about philosophical theories and their defense, not about justification.) The problem for any objective theory is that the appeal of such theories to objective likelihood seems intimately tied to the non-existence of an evil demon, for if such a demon existed it would seem that what is objectively likely to be true would be quite different than what it is (given that no evil demon exists). Hence, it would seem, evil demon worries are unanswerable on purely objective grounds alone.

It does not follow, however, that objective theories are false, nor even that we lack good philosophical grounds for accepting them. There is still the option that objective theories of justification can be derivative theories of justification, i.e., that a Cartesian-like response can be provided by some non-objective, fundamental theory of justification which in turn implies that there is an objective connection between justification and truth. The conclusion which is warranted by our discussion is not that objective theories are inadequate, but rather that if an objective account is to be acceptable, it must be accepted on the basis of a derivation as suggested above from a more fundamental theory of justification. Which theory that might be is an issue which will have to be left for another time; the point that stands, though, is that there must be one to overcome evil demon worries.

Further, seeing why the objective account is inadequate as a fundamental theory about justification shows how the circularity problem Descartes faces need not touch the account outlined above. Descartes needed the existence of God to establish that his inference to the existence of God was war-ranted; the above account need not face such a difficulty precisely because whatever account of theory construction is accepted, it must be independent of evil demon intrusions. The only question there can be here is whether any such theory is possible; I think the answer is yes, but I shall not defend that answer here.

Nor shall I be defending here that believing that God exists is justified, for that is beyond the scope of this paper as well.⁹If that belief is justified, as I think it is, we can formulate a defense of an objective connection between justification and truth, for God’s existence gives us an answer to evil demon worries. It is, after all, not possible that there is an evil demon; and, more generally, it is not possible that justification and truth diverge in a manner contrary toil a, since God exists. Thus, if one wants to know how to be a reliabilist, the answer is simple: be a theist.

III. THE DEPENDABILITY REQUIREMENTS

The other major motivation for reliabilism is the dependability motivation, of which there are two varieties: dependable persons and/or dependable procedures. The personal dependability requirement is a stronger requirement than the procedural dependability requirement. First, a person can be quite undependable and yet on a given occasion employ a procedure which is quite dependable. It is also true that a dependable person can employ an undependable procedure, yet if a person is dependable he will normally employ dependable procedures; the dependability of a procedure does not imply that it is normally used by a dependable person.

Stronger requirements run a greater risk of error, and such is the case here; I wish to argue that requiring personal dependability results in a circularity problem for defenders of personal reliability theories. Much about epistemology can be learned by considering analogous issues in ethics, so let us begin there. There is an important respect in which the moral ideal cannot be captured by any extensional account of the particular acts that a person performs, has performed, and will perform. For any extensional account of morality (an account limited in reference to acts which are actually performed) cannot account for the fact that we want persons to become disposed to do the right sorts of things. Given any extensional account of morality, it will always be possible for a person fortuitously to do all the right things, and yet be of despicable character. We want persons to be virtuous, to have character, to be disposed to do the right sorts of things; not only to do the right things.

But now consider what happens in ethics if we not only agree that persons ought to have a certain sort of character, but also insist that without such a character, they cannot do any morally right acts. Presumably, the way we develop moral character is at least partially through developing the right sorts of habits—by telling the truth repeatedly, we acquire the habit, the virtue, of being honest. But, if we defined proper behavior with theses virtues as a necessary condition for moral rightness, our theory would be viciously circular. For we would have as a necessary condition of moral rightness that virtues of a certain sort are possessed, and we would also have that, in order to develop these virtues, we must perform acts which are morally right.

There is an easy way out of this dilemma, however, for one need not accept that the only interesting question in ethics is the question “what sorts of acts ought I perform?” We can quite readily grant that there are other questions and problems worth addressing, and we can easily see that moral perfection involves quite a bit more than just doing the right things. If we allow such an extending of the sphere of ethics, we can grant that performing morally proper acts is necessary for developing moral character without making the resulting theory circular. We need only grant that, though proper moral character is a necessary part of the moral ideal, having proper moral character is not a necessary condition for performing a proper moral act.

There is much to be gained in epistemology from this description of the project of ethics. Just as we need not think in ethics that there is only one question that needs to be answered, so in epistemology we ought to refrain from thinking that the only interesting question concerns necessary and sufficient conditions for knowledge. The intellectual ideal need not be seen as being captured by any extensional account of knowing all truths, just as the moral ideal is not captured by any extensional account of doing what one morally ought to do. Just as it is important that persons develop their moral character, it is important that they develop their intellectual character. Persons ought to be disposed to follow the evidence wherever it might lead, to develop the habit of taking special precautions in areas especially vulnerable to human fallibility; we need persons who are able to clearly ascertain the implications of a body of data, and to recognize what is significant in vast bodies of information; an intellectually ideal person lacks intellectual laziness, he demonstrates creativity and imagination, and he is careful to consider all the alternatives available as well as to think up alternatives not yet known. But the importance of these characteristics should not be confused with requiring persons to have these traits in order to know, just as moral virtues are important but not necessary for proper action. We can grant that having proper intellectual character is a part of the intellectual ideal without insisting that having such character is necessary for justification.

Of course, in both ethics and epistemology, there are obvious connections between having the right sort of character and having/doing the right sorts of beliefs/actions. For having proper character entails being disposed to do the right sorts of things and believe the right sorts of things. Thus, if a person is of the right sort of character, it is more likely that he will in general do what is right and believe what is epistemically supported than if he does not have that sort of character. However, this sort of connection between proper character and the epistemically justified is only probabilistic, whereas the character motivation for reliabilism insists on a necessary contiection between the two. But. as I have argued, to insist on a necessary relation makes one’s account circular; for one develops a certain sort of character at least partially by performing the right sorts of acts and by following the evidence where it leads. Insisting on the necessary connection between the two involves a certain sort of tunnel vision: thinking that the intellectual or moral ideal is achieved when a person does all and only the right acts or when all of his beliefs count as knowledge. There is no good reason to think that that is all there is to either ideal, and there is plenty of reason to think otherwise.

Not distinguishing the cognitive ideal from what is necessary for justification causes improper evaluations of particular cases. For example, consider a gypsy lawyer who is the lawyer of a client accused of eight murders. The client admits having killed the first seven but denies having murdered the eighth. Nonetheless, everyone in the community, including the lawyer, is convinced that he is guilty of all eight. As gypsies do, though, the lawyer consults the tarot cards, which say that his client is innocent of the eighth murder. His gypsy nature then causes him to believe what the tarot cards say, and what the tarot cards say causes the lawyer to reexamine the evidence. Upon re-examining the evidence, however, the lawyer clearly recognizes that the evidence, rather than confirming his client’s guilt, completely exonerates him of the eighth murder. But, being of rather impressionable character, the lawyer recognizes that if it were not for the influence of the tarot cards, he would be unable to avoid being swayed by public opinion both about what the evidence shows and about the guilt of his client. Thus, the tarot cards are solely responsible both for the origination and sustenance of his belief that his client is innocent.’⁰

Audi claims that, though there is a justification for his belief, the lawyer’s believing is not justified (contrary to Lehrer’s claims):

“(G)iven his faith in the cards, he would have believed p even if it had been false, indeed, even if, on the basis of the cards, it had not been rendered so much as objectively likely (to any degree) to be true, i.e., very roughly, likely to some degree given the actual facts relevant to p.... (This point) strongly suggest(s) that S does not justifiably believe p. It is, after all, simply good fortune (because the cards happened to be right) that S did not believe something false in place of p. Surely if one’s belief that p is justified by good evidence, it cannot be simply good fortune that one did not believe something false instead.”’

Audi claims that a person cannot justifiably believe a proposition if he just happens to be lucky in having a belief for which he has adequate evidence. Yet, whether a person is lucky or not is a matter of what sort of character that person has; it is a matter of whether that person would believe the truth had things been a little different. If Audi wishes to hold that proper character is necessary for justification, he must tell us how we acquire such character without appeal to forming justified beliefs.

Further, there are strong intuitive grounds for rejecting Audi’s account of this case. What we ought to ask is whether the gypsy lawyer knows that his client is innocent of the eighth murder, and here I think the answer is clear: of course he knows that his client is innocent (note what the lawyer could legitimately claim in response to the amazement of the town at finding out the truth—he could quite rightly say, “I knew it all along”). If we must go about purifying our motives before we can ever know anything, most of us would know quite little, just as we could not do much that is right if we were required to purify our motives in order to perform a morally proper act. We can do the right sorts of things for the wrong motives, and we can have knowledge even though what sustains our belief is not what allows us to know.

What we immediately notice about the gypsy lawyer is that something has gone wrong in his cognitive life; but imputing this cognitive fault against his justified belief only results by confusing justification with intellectual ideality. Just as we can perfectly well grant that someone did the right sort of action even though his character was somewhat marred (think of cases in which one recognizes in oneself improper motives, but one decides that one ought to do what one is improperly motivated to do anyway), one can have knowledge in a certain case without requiring that one’s intellectual character be spotless. The lawyer may even regret his inability to break his trait of trusting the tarot cards; nonetheless, he knows in this case that what the tarot cards led him to believe and sustain him in believing is true. Though he is in a less than intellectually ideal relation to the proposition in question, the lack of ideality concerns his intellectual virtues, not whether he has knowledge or not.

So we can see that personal reliabilism has problems if the intuition that we want persons to be reliable is taken to be a necessary conditon for epistemic justification. But what of the weaker claim that reliable procedures are necessary for epistemic justification? Don’t we have here a necessary condition for epistemic justification, and thus isn’t some form of reliabilism true?

Procedural reliabilists deny that what a person would believe in close counterfactual circumstances is relevant to the justificatory status of the present belief. Counterfactual cases are relevant, according to procedural reliabilists, but it is counterfactual situations about the procedure, not the person, that are important. One way to understand this position is to see it as a retreat from personal reliabilism in the face of the difficulties above. Since we cannot coherently require that a person have a proper cognitive character in order to have a justified belief, we must retreat to a more minimal claim that the person must instance (in the actual circumstance) what would be proper character if the person in question were usually to proceed in that fashion. The requirement that the person normally proceed in that fashion is dropped; what remains is the requirement that the procedure be one that a person of good cognitive character would employ.

When understood in this way, though, it is far from obvious why a procedural requirement is necessary for justified belief. We saw earlier that proper cognitive or intellectual character is part of the intellectual ideal; hence, so is exhibiting such a character in a given circumstance. Yet, if having a proper intellectual character is not necessary for justification, why should exhibiting it in a given circumstance be? To appreciate my puzzlement, recall that central to the procedural requirement is temporal duration. To require that proper procedure is followed allows that facts solely about the past can imply that one’s present belief is unjustified. Again, consider ethical analogies. One’s past conduct can affect which options are presently available to one, but which of those options is preferable is a matter of their present moral worth. Suppose a person has a thoroughly despicable past, but suddenly has a change of heart and begins to act differently. Clearly, some actions he could have done had he behaved differently in the past he will not be able to do; perhaps if he had been more careful in the past, he would not be so sick that he can do nothing but lie in bed and recover while his family suffers because of his inability to work. Yet, we should not say that he is now behaving improperly by not getting out of bed and going to work. What he should do now depends on what he is capable of doing now, and which of the available options is presently superior on moral grounds.

The same point holds for epistemic matters as well. We may follow procedures that damage our overall chances of finding the truth and avoiding error; yet what is relevant at a particular time is not these overall chances but rather what the present circumstances indicate. Even in inductive cases, where the past seems to be relevant to present justification, it is not strictly the past that is relevant, but rather our present memory of past cases like the present one that determines the justificatory status of belief (or perhaps the present fact that we should recall and take into account certain information about the past that makes a belief less than rational). The gypsy lawyer case is as instructive here as it was with regard to a character requirement. The procedure which produces and sustains his belief (namely, reading the tarot cards) is clearly unreliable, yet he has knowledge. How? Presumably in virtue of his (present) awareness of the evidence and what it shows.

The lesson here is simple. If having a proper intellectual character is not required to have a justified belief, there is no reason to think that following a procedure that would be followed by one with a proper intellectual character is required either. One’s procedures prior to t are simply not relevant to what the circumstances at t show to be true or false; it is what the circumstances at t show that is relevant to what one ought to believe.

One way to express the lessons of cases like the gypsy lawyer case employs the notion of epistemic luck. Procedural reliabilists grant that some epistemic luck is possible, for a person can just happen to employ proper procedure on a given occasion and yet have knowledge. In. allowing this sort of luck, procedural reliabilism avoids the circularity problem faced by personal reliabilism. What gypsy-lawyer-like cases show is that an even greater sort of epistemic luck is possible: a person can have a belief produced and sustained by a faulty procedure which just happens to be an instance of knowledge as well. The distinction between the intellectual ideal and justification provides an explanation for this phenomenon of epistemic luck.¹²

We have been thinking of procedural reliabilism as a sort of limiting case of personal reliabilism: the procedure is proper if and only if it is one that a reliable person would use. Yet, we can now see that even if we think of the procedural requirement as completely independent of the character requirement, there is no reason to take it to be necessary for justification. Again, it is clearly part of the intellectual ideal that we follow proper procedures; I tend to think that this is so because it is part of the intellectual ideal that we have proper cognitive characters. But even if the procedural requirement is independent of the character requirement, its being a part of the intellectual ideal should not prompt us to think that it is a necessary condition for justification. The past is irrelevant to what is presently shown to be true.

IV. CONCLUSION

Both the truth motivation and the dependability motivation deserve a place in any acceptable epistemology; the difficulty with reliabilism is not knowing where to put the elements that satisfy these motivations. The truth motivation can, at best, be satisfied by a more fundamental theory of justification than reliabilism, for directly satisfying that motivation results in a theory burdended by counterexamples. In addition, epistemological questions are questions regarding the cognitive or intellectual ideal, and the dependability motivation for reliabilism is clearly an important motivation to satisfy in any adequate account of the cognitive ideal. Developing a theory of justification, however, which satisfies the dependability motivation results from confusing whether a person is cognitively ideal with whether a person has a justified belief.

A complete epistemology will include elements which satisfy the motivations for reliabilism; however, the way in which these motivations are satisfied will be quite different than reliabilists have imagined. Put quite negatively, in order to satisfy reliabilist motivations, one ought not to be a reliabilist at all.