|Find Enlightenment||Contents||Previous Chapter|
IS EPISTEMIC CIRCULARITY VICIOUS?
I. What's Wrong with Self-Support Meta-Foundationalism?
Alston and others deny that it is necessary, to be justified in believing a justification standard on the basis of an argument, to adduce or even be aware of the licensing j-standards for the premises of the argument. They claim, moreover, that it is perfectly possible to be justified in believing such an argument's premises without even being aware that that in virtue of which we are justified in believing the premises of the argument is the conclusion itself. Such considerations they cite as making self-supportism plausible.
Thus, in the sense defined in Chapter 1, a standard might support itself: the standard might license an argument containing that same standard as the conclusion. But such an argument is, as I pointed out in Chapter 1, epistemically circular. Hence Alston and other defenders of self-supportism are concerned to argue that epistemic circularity is not vicious. If they are correct, then self-supportism has an excellent chance of being the proper solution to the Problem of Meta-Justification.
Robert Nozick is a self-supportist who has clearly seen this circularity and embraced it. His attitude is typical. Speaking of a doxastic system containing epistemically circular arguments, he writes:
Looking at the overall structure, however, we seek just such [epistemic] "circles". It is not surprising that some features objectionable in intermediate connecting links within a structure might be desirable in fundamental principles that underlie the whole structure. Local vices, global virtues. It is desired that there be principles which underlie and yield the rest, which subsume themselves and so do not dangle or lead to infinite regress. The discovering and uncovering of such fundamental truths is not a crisis or trauma for philosophy, but a triumph.
Nozick does not give any reasons at all to think that the "local vice" of epistemic circularity is a "global virtue" when embedded in a "global" system's foundations. He simply states the claim baldly, though eloquently, with the appearance of having given an argument for the view. I am reminded of G. E. Moore's so-called "Open Question Argument" against naturalism on that score; practically everyone admits that the argument is no good, but they continue to say that the argument makes the trouble with naturalism intuitive. Similarly, declarations such as Nozick's have, no doubt, made it seem plausible that epistemic circularity is benign, without having given any good reasons for thinking so.
The plan of this chapter is simple. First, I will use Alston's view to introduce the self-support meta-foundationalist's stance toward epistemic circularity. Second, I will interpret and evaluate arguments, by Braithwaite, Black, and Van Cleve, that the circularity is benign. I will conclude that none of those arguments do any more than shift the burden of proof to the person who thinks that the circularity is vicious. And so, third, I will meet that burden of proof, by presenting four arguments that the circularity is vicious. I will conclude that self-supportism must be rejected.
II. Alston's View as an Introduction.
On Alston's view, epistemic circularity is a puzzling difficulty, but it is not vicious in the sense defined in Chapter 1. He says this in several places, but his most complete discussion of this point is found in his essay "Epistemic Circularity." What Alston describes as a "track record" argument for the reliability of sense-perception has as its conclusion (Alston's numbering) (II), while (V) is admittedly presupposed by the argument's premises. Quoted from Alston's essay:
(II) Sense experience is a reliable source of perceptual beliefs.
(V) If one believes that p on the basis of its sensorily appearing to one that p, and one has no overriding reasons to the contrary, one is justified in believing that p.
Alston proceeds as follows. Consider a track record argument roughly of the form of the second from Chapter 1, Section II above, though applied specifically to sense perception. Now, (V) resembles a generalized form of what I was calling a "bridge principle" in Chapter 1. According to Alston and other reliabilists, something like (V) needs only to be true in order for one to be justified in various perceptual beliefs (premises of the form (nb); see Chapter 1, Section III). It is crucial to realize, for Alston, that being justified in believing all the requisite (nb)'s in a track record argument does not require being justified in believing (V); it requires only that (V) be true. "So what's all the fuss?" one might say. "Sure, we can see that track record arguments are circular; we might find them unpersuasive for that reason. But their unpersuasiveness to us, happily, does not mean that one could not, in theory, get a justified belief in (II) using a track record argument. So their epistemic circularity is not vicious."
In order to make this argument go through, Alston has to talk about the conditions under which one can get a justified belief in a conclusion based on an argument for it. But we may put aside that talk, I think, and concede that, at least at first glance, his analysis looks right. As I would put it: we certainly do not need to be justified in believing j-standards in order to have beliefs justified in accordance with those very j-standards. We can be justified in holding premises such as (nb) to be correct, while being ignorant of some things in virtue of which we are so justified. We can even be justified Alston goes on in believing the conclusion on the basis of such premises, even though the truth of the conclusion is a necessary condition of our being justified in believing the premises.
This is (as we shall see) similar to the solution to the problem of induction given by Max Black, James Van Cleve, and others who argue for inductive justifications of induction. It is possible to use induction to argue that induction is reliable, without saying that we must first, as a precondition of our using such an argument, know that induction is reliable. In everyday life and in science, it is no criticism of inductive practices that we lack a justification of those practices; why should it be any different, then, when arguing for induction?
These sorts of moves are understandably persuasive or perhaps "seductive" would be a better word. For all their persuasiveness, they do not seem to take full account of the problem. As Alston himself puts it, "This is to offer stone instead of bread." A track record argument can justify belief in its conclusion only if its conclusion is true. If we are not simply going through the pointless exercise of reaffirming what we already consider obvious, it will not satisfy us to give an argument for a conclusion when we are aware that in making the argument we are taking the conclusion for granted.
Alston has an excellent way of bringing out this point. He says that the reliability of crystal ball gazing, for example, might well turn out to be defensible if it is reliable. Similarly for any sort of strange, occult practices. But if we took it for granted that crystal ball gazing were reliable, and then produced a track record argument showing its success, no one would be impressed. Why, then, should we be impressed by epistemically circular arguments for the reliability of sense perception, or for some principle of induction?
Throughout The Reliability of Sense Perception, Alston insists that epistemic circularity poses a deep and difficult problem. But Alston's position there was curious. After everything he said about what a problem it is, he still maintained the position that we can get a justified belief from a circular track record argument. If all we wanted was to justify our beliefs, Alston said we can do that with an epistemically circular argument; it does not even matter that we know that we are presupposing the conclusion.
So perhaps it should not be surprising that Alston reversed his view more recently: "I now think that I overreacted to the problem of epistemic circularity.... [I]f we are going to rely on an epistemically circular argument anyway, we might as well go with the more ambitious ones that seek to show it to be true that SP [the belief-forming practices associated with sense perception] is reliable." So in his most recent published writing on the subject, he appears to hold that epistemic circularity is not such a problem after all. But what about the fact that epistemically circular track record arguments are equally available for all sorts of disreputable practices, such as crystal ball gazing?
Those practices, Alston says, might not display significant self-support . Sense-perception gives us information about the physical make-up of the senses, which information allows us to fine-tune our accounts of why and when sense-perception is reliable; and similarly for other natural cognitive processes such as memory and reasoning. But this sort of self-support cannot be assumed to be a feature of occult practices.
Alston now apparently regards his trenchant analysis in The Reliability of Sense Perception showing how attempt after attempt to argue for basic j-standards runs afoul of epistemic circularity was beside the point. If there is nothing wrong with epistemic circularity per se, then there was no point in showing how all those arguments were epistemically circular.
I completely disagree. Alston, Black, Van Cleve, and all the rest have made an important realization that justified belief in the premises does not necessarily require justified belief in what the argument presupposes but have incorrectly inferred that epistemic circularity is benign, and hence entirely underestimated the size of the problem.
III. Arguments for Benignity.
Other writers, including R. B. Braithwaite and Max Black, have promoted self-supportist solutions to the problem of induction, as has, more recently, James Van Cleve; Van Cleve also has a self-supportist way out of the Cartesian circle. These writings are generally directed to the problem of induction rather than epistemic circularity, but it is not difficult to see their relevance. It is worth investigating whether their arguments might be converted into good arguments that epistemic circularity is not vicious.
So let us begin with R. B. Braithwaite, who in his 1946 Tarner Lectures proposed that inductive arguments for inductive policies are not viciously circular. He describes inductive reasoning as "the use of inductive policies," where "inductive policies" are "policies for establishing general hypotheses in accordance with inductive principles of inference on the basis of empirical data"; "inductive principles," in turn, are said to be "those discussed in books on inductive logic and scientific methodology," which include "simple enumeration" and "principles of elimination."
Following Peirce, Braithwaite claims that inductive policies are to be justified by pointing out their effectiveness (in the past); this he calls a predictionist justification of induction. He claims that when a policy satisfies a certain (rather complicated) criterion of effectiveness, it is justified; in that case, inferences made in accordance with the policy are "valid," and beliefs in the inferred conclusions (under certain conditions) are "reasonable." Without explicating the entire criterion of effectiveness (which is unnecessary for our purposes), it may be adequately summarized as follows: to be effective, a policy must be such that, in a given time period, "many" of the hypotheses established by its use both (1) have not been empirically refuted, and (2) have been empirically confirmed at least once.
As Braithwaite himself points out, many will identify a circularity in an attempt to justify an inductive policy by pointing to its effectiveness. To be justified in believing the premise that a policy is effective evidently requires another inductive argument. If the inductive argument licensing belief in the premise depends on a different inductive policy, there is no circularity; but the same problem of justification arises for that policy. Hence either we have an infinite regress of inductive policies, or we shall have to say that an inductive policy establishes its own effectiveness which at first glance appears to be viciously circular. If we nix the infinite regress option, the predictionist justification of induction presupposes the validity of induction by simple enumeration, and "the validity of induction by simple enumeration presupposes its own validity; and this, it is alleged, is a viciously circular justification for induction."
Braithwaite's reply to this problem predates Alston's stance toward epistemic circularity by about forty years. "The first move in the rebuttal" of the circularity charge, Braithwaite writes, "is that the proposition `presupposed' in the predictionist justification of an inductive inference does not function in the inference as an additional premise." In other words, when we argue for the effectiveness of an inductive policy by using simple enumeration, no rule of simple enumeration operates as a premise in the argument. To say that it does immediately raises a different, but very familiar, infinite regress. So the circularity here is not premise circularity.
"I do not wish to deny that there is a sort of circularity" in this case, Braithwaite says, "but it is a peculiar sort of circularity whose viciousness is by no means obvious." Braithwaite goes on to enumerate a number of different criteria for the reasonableness of an inferred belief q, and reduces the possibilities to three (p is a single conjunctive premise of the argument, and r is the rule licensing this particular inference):
VI B reasonably believes p and believes r.
VIII B reasonably believes p, and r is true.
IX B reasonably believes p and believes r, and r is true.
Braithwaite says that arguments for the effectiveness of an inductive policy that fit these criteria for a person "will be valid without any circularity" by which he evidently means premise circularity. (Later he stresses, "In none of these three cases is there any vicious circularity.")
Braithwaite appears to take the considerations described in the foregoing three paragraphs as decisive. Having established that the inference in question is not premise circular, he says that in the case of VIII in particular, it is possible for the inferred belief to be reasonable by these criteria, "whether or not the inferrer knows or believes" that the inference policy he uses "is effective or indeed whether he considers the question of its effectiveness at all." Braithwaite admits that an inductive argument for the effectiveness of an inductive policy has an "implicit circularity"; but this is no trouble, because "the implicit circularity only arises from the inference-machine becoming self-conscious about the way in which it operates."
Of course, this is very similar to what Alston has said more recently about epistemic circularity. But it is hasty and unwarranted to conclude on such grounds alone that epistemic circularity is not vicious. It is surely interesting and important to recognize that epistemic circularity does not entail the more obviously vicious premise circularity and that one need not even be aware of the rule that licenses an inference that justifies a given belief. But it might be the case and is in fact that the reason epistemic circularity is vicious has nothing to do either with premise circularity or with awareness of a licensing inference rule. That is what I will argue in the sections following the present one.
Of all the philosophers I will discuss in this section, Max Black is most famous for arguing, in a series of articles in the 1950's and'60's, that inductive arguments for induction are not vicious. His argument differs slightly from Braithwaite's.
Black presents two so-called "self-supporting" arguments for (admittedly idealized) inductive rules. ("Self-supporting" is a surprisingly apt name given my label for the move in the MRA under examination in this chapter.) For example:
(a1): All examined instances of the use of R1 in arguments with true premises have been instances in which R1 is successful.
All instances of the use of R1 in arguments with true premises are instances in which R1 is successful.
R1 is this rule: "To argue from All examined instances of A's have been B to All A's are B." Black points out that this argument has been "so formulated that (a1) is governed by R1." In other words, R1 is supposed to be the inference rule that licenses the inference from the premise to the conclusion of (a1). (That is easy to see: replace `A' in a1 with `examined instances of the use of R1 in arguments with true premises' and replace `B' with `instances in which R1 is successful'.) But the conclusion is tantamount to the claim that R1 is "reliable" (as Black says). And so we have an argument such that its conclusion claims that its own licensing inference rule is reliable.
Black proceeds to argue that there is no sense in which this argument is "guilty of circularity." He says that there are two ways that an argument can be circular: first, what we earlier called "premise circularity," and second, the situation that obtains when "at least one of the premises is such that it is impossible to get to know its truth without simultaneously or antecedently getting to know the truth of the conclusion." (The latter situation, if not coextensive with epistemic circularity by my definition, clearly is a variety of epistemic circularity.) Black spends considerable time arguing (successfully) that (a1) is not circular in either of these senses. He appears content to think that this disposes of any concerns about any vicious circularity.
Black asks when the use of an inductive rule R1 might be legitimate; to sum up his discussion, he says that the most stringent requirements will have us subject R1 to searching criticism. If R1 survives such criticism, its use is, certainly, legitimate. Black then asks, "Can a self-supporting argument be correct without triviality?" He appears to admit that (a1) might well be trivially self-supporting:
Well, no doubt, inductive inferences will have been used in the course of finding good reasons for one's confidence in the reliability of R1... If that very same inductive evidence for R1's reliability is now produced again as the premise of (a1), (a1) will indeed yield no new knowledge, and will then, indeed, lead to useless repetition of what was known at the outset.
The point is that if, when R1 was subjected to criticism, we produced a lot of evidence that is now adduced in the premise of (a1), then (a1) will be trivial. But he goes on to say (his italics), "[T]here is nothing in the specification of (a1) that requires us to confine ourselves to the evidence that we previously had for R1's reliability." We may, then, supplement the evidence we previously had for R1's reliability with new evidence, and this will strengthen our argument for the conclusion of (a1). Hence our argument might be self-supporting but not trivially so. If Black is right, we may use self-supporting (rule circular ) arguments to augment the probability of an inductive rule whose basic reliability is already established.
That is not a claim that we need evaluate here. Even if true, it does not bear on how we could, without triviality (or vicious circularity), establish the basic reliability of an inference rule in the first place. No doubt some sort of bootstrapping technique is unobjectionable, once a weak version of a rule is on the table. But that does not impugn or otherwise bear on the criticisms we have in store for epistemic circularity.
So, Black, like Braithwaite, points out that self-supporting inductive arguments are not premise circular; Black adds that we may use such self-supporting arguments to bootstrap our way to stronger rules. As we shall see, that is not enough to establish the virtuousness of epistemic circularity.
We have examined various attempts to mitigate the viciousness of epistemic circularity; these attempts make a variety of quite legitimate claims, but none of them actually shows that epistemic circularity is not vicious. Braithwaite is content to point out that the "implicit circularity" that is admittedly part of inductive arguments for induction is not premise circularity and thus nothing to worry about. Black makes a different point, saying in effect that, given a weak induction rule, one can bootstrap one's way up to stronger versions; that might be perfectly correct. Alston and Van Cleve  point out that one need not be aware of an argument's licensing inference rule, or of what justified belief in the premises presupposes, when the inference rule, or a presupposition, is the argument's conclusion. But in none of these cases has anyone actually succeeded in giving a positive argument that epistemic circularity is benign.
Perhaps we shouldn't expect anyone to do this; proving a negative is often very difficult. But clearly the burden of proof has been shifted onto those who think that epistemic circularity is vicious. Several leading twentieth-century philosophers advanced various views on why epistemic circularity might be vicious, and they successfully showed that it is not vicious for any of those reasons. It is now up to their opponents to show why it might be vicious, if they still think it is.
So I will attempt to meet this challenge. Many people who come across epistemically circular arguments feel that something is seriously amiss in them. The following four arguments should locate exactly the source of this intuition. To begin, then, with the Argument from Philosophical Requirements.
IV. The Argument from Philosophical Requirements.
It is possible to be justified in believing the premises of a track record argument without that belief being consciously based, for its justification, on the j-standard stated in the conclusion. This point I concede. But I do not concede that it is possible to satisfy ordinary philosophical requirements on being justified in believing a philosophical claim on the basis of an argument, if that argument is epistemically circular. Let me explain.
In attempting to demonstrate the viciousness of epistemic circularity, we might seriously apply remarks in defense of such circularity to actual attempts to justify actual standards. Will such an approach actually wash?
Philosophers, inquisitive as they are, look into how various claims can be justified, what inferentially depends on what, and so forth. If they notice (as they often do) that one cannot be justified in believing the premises of a certain argument, unless another, perhaps questionable proposition is true, they are apt to describe this state of affairs:
(JP) One can be justified in believing a certain set of premises (and hence, the conclusion on the basis of these premises), only if a given proposition is true.
It is quite natural for philosophers to investigate whether such a necessary condition actually holds that is, whether the proposition in question is true by bringing forth various considerations for and against it. This happens a lot in discussions of a wide variety of philosophical topics. Of course, it does not happen always with every such necessary condition, as I will discuss below.
When we come across Alston on epistemic circularity (and others writing on induction or Descartes' circle), we encounter a curious phenomenon. Alston does not give a track record argument for the reliability of sense-perception; he only talks about such arguments. Well, suppose that he has given a full-blown track record argument for the reliability of sense-perception, in addition to the rest of his philosophical work. But he does not bother to investigate whether a certain proposition, on which the argument depends, actually is true. So even though he admits that the proposition must be true if the argument is to succeed, he fails to argue for the proposition.
In our fictional Alston's defense, one might point out that a track record argument itself constitutes sufficient investigation of the truth of its conclusion and hence of the proposition on which the argument depends. But in any other philosophical context, the latter move would be regarded as obviously question-begging and hence illegitimate. What is in question is whether a track record argument for a j-standard about sense-perception actually can be used to show belief in the standard to be justified. We discover and are now well aware that track record argument for the standard in question is successful in this regard only if the conclusion is true. Hence it is natural for us (at least, in any other philosophical context) to require some entirely separate argument for the conclusion. When our fictional Alston does not fulfill this requirement, he shirks his intellectual responsibilities.
It does not help to say that the actual Alston merely discusses epistemically circular arguments, and does not offer such an argument himself; that's irrelevant to my point. Hardly anyone ever actually gives the sort of circular arguments they're fond of talking about (except, perhaps, Descartes and a very few others). Avoiding the attempt makes it easy to dodge the difficult questions they would face if they made the attempt difficult questions that Descartes did famously face. What Alston says about epistemically circular arguments is that an argument that presupposes its own conclusion can supply ample justification of our belief in that presupposition. My claim is that if such an actual argument were to be made in (what else?) a philosophical context, no one would buy it. And rightly so.
Nor does it help here to say that we can be justified in believing the premises without being antecedently justified in believing the conclusion. For, in a philosophical context, it is usually the case that we naturally and properly require that, if the premises are justified by any further considerations, those further considerations actually be adduced. When we discover that the conclusion is among the considerations essential to justifying the premises, we do, and should, throw the argument out as ill-suited to its purpose.
To put my objection in different words. Alston and his like argue, in essence:
(i) If S is justified in believing the premises of epistemically circular argument A for justification standard J, then (provided other conditions are met) S is justified in believing J on the basis of A.
(ii) S is justified in believing the premises of argument A for justification standard J.
(iii) Hence, S is justified in believing J on the basis of A.
Suppose S is our fictional Alston; Alston has given track record argument A. We charge him with begging the question. In his defense, he states (i)-(iii): a meta-justification. Now, (i)-(iii) is a metalevel argument, made by a philosopher (Alston ) in defense of a philosophical claim (that Alston is justified in believing J on the basis of A). In this context, we may ask: What evidence can be adduced in support of (ii)? Is it simply being asserted without any argument at all?
Here Alston faces a dilemma. Suppose the answer is "Yes." So he thinks he can assert that he's justified in believing the premises of A without giving any argument for that assertion. Notice, the question for Alston is not whether he is justified in believing each (nb) without arguing for each (nb). The question is: In a philosophical context, can Alston legitimately assert, without argument, the following metalevel claim: "I am justified in believing each (nb) (along with the track record argument's other premises)"?
It seems not, particularly when we can see that justified belief in the premises requires the truth of the conclusion Alston wishes to draw. If we permit such a move, there is no good reason why Alston should not be allowed simply to assume that the conclusion is true and forego arguing altogether.
Suppose now that the answer is "No" that (ii) is asserted with some argument. Then I suppose Alston's argument would be roughly like this: "We have reliable belief-forming practices that result in beliefs stated in the premises of A; hence, those beliefs are justified." In that case, Alston never had any point in making the track record argument at all. He might as well have asserted J without argument, since he was going to do so eventually anyway (at least, after heavy questioning).
It is easy to see that we have forced the fictional Alston, who before gave an argument that was merely epistemically circular, to a much more obviously vicious circularity. If the j-standard J can be stated roughly as that we do in fact have reliable belief-forming processes (of some particular sort), then when Alston offers the argument in favor of (ii), including J as a premise, then he obviously begs the question. We may throw his argument out.
Perhaps this is a bit too fast. Notice that when Alston makes the argument in favor of (ii), which includes J as a premise, the ultimate conclusion of his argument is not J but instead (iii), i.e., that Alston is justified in believing J on the basis of A. That is not premise circularity.
The problem quickly resurfaces. If Alston is to be justified in believing (iii) on the basis of this philosophical argument (i)-(iii), and he is justified in believing premise (ii) in part on the basis of J, then he must be justified, antecedently, in believing J. In order to be situated properly to justify his belief that S is justified in believing J on the basis of A, Alston must already be justified in believing that J.
And philosophers, being inquisitive as I said, will of course want to know what justifies Alston in believing that J. As we assumed (in this section, anyway) at the outset, the best that Alston can do is to give a track record argument (that meets certain conditions): something like A. And now we have come around full circle.
It is not as though Alston does not know he is in this situation described by the present line of reasoning. In Alston's own terms, all that I have shown is that no belief in any j-standard can be "fully reflectively justified" (FRJ ). Alston explains: "When a belief has been FRJ, no questions are left over as to whether the subject is justified in accepting some premise that is used at some stage of the justification." A belief is FRJ only if it has been successfully supported by an argument, and each premise of that argument has in turn been successfully supported, and so on. Obviously, barring premise circularity, FRJ is impossible.
In my comments about what philosophers properly require of themselves and each other, I did not wish to imply that it is always required, for each one of their claims, that some argument be given for that claim. There are, because there must be, exceptions.
It is possible to give a persuasive argument that there must be exceptions. Assume first that ought implies can: if I ought to do something, in any plausible sense of `ought', then there is some sense in which I am (or, perhaps, was) able to do what I ought. Contraposing: if there is no sense in which I am (or was ever) able to do something, it is not the case that I ought to do it. Apply that insight to the case of philosophical argumentation. Surely it is impossible to "argue for everything"; hence it is not the case that one ought to argue for everything.
After having made this concession, can I still maintain, against Alston, that epistemic circularity is vicious? Perhaps not on these grounds. The Argument from Philosophical Requirements depends on the premise that philosophers do rightly require of each other that they adduce their justification when they can say what it is. And I have come to the conclusion that when philosophers do this, they discover that when they broach the subject of epistemic circularity, they run in a fairly tight circle. But they must be permitted to say something without argument. Why not, then, (i) and (ii)? Or better yet, the premises of A?
More generally, where, in this circle, should one begin? The answer is not clear. It is not clear what the proper starting-points in this area of philosophy are. By this talk of starting-points, I mean starting-points in philosophical argumentation; hence the items constituting starting-points are premises of philosophical arguments, or claims that support other claims but that are not given any support themselves. A proper starting-point would be, then, some statement, made in the course of doing philosophy, that is neither given any support by any other statement nor properly regarded as in need of any support. Evidently, what makes the notion of "proper starting-point" so obscure is this: it is unclear what, if anything, makes a statement properly regarded as in need of argumentative support.
This is not an issue that I can properly adjudicate here, however interesting and important it might be. I can observe, however, that Alston's defense of epistemic circularity as (in at least some cases) virtuous implies that he believes the premises of epistemically circular arguments are among the proper starting-points of philosophy. Is there anything wrong with that?
I think one point can be urged against it. If indeed there is a fairly tight circle in which the relevant premises and standards surrounding these meta-epistemological issues can be found, and essentially we must choose somewhere to begin in the circle, then why choose to assume the premises of an epistemically circular argument? Why not, instead, simply assume that the certain standards are correct, or that sense perception and other basic ways we have of gathering knowledge are reliable? We are, admittedly, assuming such things in saying that we can have a justified belief in the conclusion of epistemically circular arguments. So what possible motivation can we have for failing to make the assumptions themselves the "starting points"?
The motivation is obvious. Those who want to depend on epistemically circular arguments to support their j-standards and their claims about the reliability of sense-perception (etc.) want to give such conclusions the appearance of support that an argument can give. To make the conclusions on which claims to having justified beliefs from such arguments depend into philosophical starting-points appears to be so much question-begging. So the defenders of epistemic circularity want to argue for such starting-points, if only to remove some of the appearance of begging the question but only some of the appearance, because after all, admittedly, knowledge claims based on epistemically circular arguments do in fact beg the question just as egregiously.
So, I advocate (in Chapter 4) simplifying the situation by openly assuming what will have to be assumed in any case.
This does not, admittedly, by itself constitute a knock-down argument that epistemic circularity is vicious. The Argument from Philosophical Requirements at a minimum establishes that we shall have to choose some philosophical starting-points from among a limited set of propositions. We can choose some and build question-begging arguments for the others; or we can simply choose as starting points those others.
The latter choice seems most sensible, because it avoids begging the question. In this way we undercut the motivation for relying on epistemically circular arguments and, hence, for maintaining that they are not always vicious.
V. The Equal Availability Argument.
Consider what Alston says about epistemic circularity in, for example, The Reliability of Sense Perception. He does not say that it is vicious. "Epistemic circularity does not in and of itself disqualify the argument," he says. Then he goes on,
But even granting that point, the argument will not do its job unless we are justified in accepting its premises; and that is the case only if sense perception is in fact reliable. This is to offer stone instead of bread. We can say the same of any belief-forming practice whatever, no matter how disreputable.
The trouble, then, as we discussed in Section II above, is that epistemically circular arguments can be given for both good and bad practices, and hence such arguments cannot be used to distinguish reputable from disreputable practices or true from spurious j-standards. This strikes me as an excellent reason to believe epistemic circularity as such is vicious; but Alston himself was initially unclear on that point. In fact, Alston's earlier position on the viciousness of epistemic circularity is rather difficult to pin down.
Alston said (as quoted above) that circularity "in and of itself" does not disqualify an argument (apparently for the purpose of getting justified beliefs). But he also says further down on the same page, "Hence I shall disqualify epistemically circular arguments on the grounds that they do not serve to discriminate between reliable and unreliable doxastic practices." So circularity "in and of itself" is not disqualifying, but the fact that circular arguments can be made for reputable and disreputable practices alike is enough to disqualify those arguments. This is, at best, a puzzling conjunction of claims:
(1) Epistemic circularity, "in and of itself," does not disqualify an argument.
(2) Epistemically circular arguments are equally available for reputable and disreputable practices alike; hence, all epistemically circular arguments are disqualified (for that reason).
Recall now that Alston has modified his position, so that according to his most recent article on the subject, he is now willing to accept some epistemically circular arguments in order (for example) to justify belief in the reliability of sense-perception and other practices that display "significant self-support" (henceforth, in this section, `self-support'). So it appears he now rejects (2). Alston still maintains, however, that those circular track record arguments for the reliability of practices that do not display self-support, such as crystal ball-gazing the "mere track record arguments" (MTRAs) are "under the ban of being equally available for any doxastic practice whatever, no matter how disreputable."
So Alston now advocates (1) together with a modified version of (2):
(3) Epistemically circular track record arguments are equally available for reputable and disreputable practices alike; hence, epistemically circular track record arguments for doxastic practices that lack significant self-support are disqualified (for that reason).
To make sense of this modified view, it is important to bear in mind that Alston wants (contrary to his original aim) to rely on epistemically circular arguments to show that sense perceptual doxastic practice (SP) is reliable. So he has to produce a way to distinguish two varieties of epistemically circular arguments, to wit, the virtuous and the vicious.
He proposes a way to make the distinction: the arguments for the doxastic practices that are self-supporting can be virtuous, while the arguments for the doxastic practices that are not self-supporting are vicious. The former arguments are not merely track record arguments; apparently, they exhibit the fact that a doxastic practice is self-supporting. The latter arguments, on the other hand, are "mere track record arguments" and do not exhibit a practice's self-support.
Why, I wonder, are the allegedly vicious arguments held to be vicious at all? Alston's stated view on this is that such MTRAs are under a "ban of being equally available for any doxastic practice whatever."
But this is very puzzling. After all, why is it that MTRAs are indeed equally available for the unacceptable practices? It is precisely because such arguments are epistemically circular. It isn't because they are track record arguments and it isn't because they do not exhibit a doxastic practice's self-support. As anyone familiar with the phenomenon of epistemic circularity knows very well, it is epistemic circularity itself that explains why MTRAs are equally available for unacceptable practices. Hence, if it is the equal availability of mere track record arguments that makes them vicious, and it is epistemic circularity that makes them equally available, then (it seems reasonable to suppose) epistemic circularity per se is vicious. This argument is worth spelling out a bit more carefully:
(i) If a type of argument is equally available for patently unreliable doxastic practices, then that type of argument is unacceptable.
(ii) Since MTRAs are equally available for patently unreliable doxastic practices, they are unacceptable.
(iii) But it is epistemic circularity that makes MTRAs equally available for patently unreliable doxastic practices.
(iv) Therefore (probably?) all epistemically circular arguments are unacceptable (vicious ).
Admittedly, the conclusion does not follow deductively. Perhaps, as Alston's defenders will want to suggest, I am missing the point. The point is that arguments that exhibit a doxastic practice's significant self-support are different from MTRAs precisely because they exhibit that self-support. Objectionable practices cannot be given such arguments; so, such arguments are not equally available for objectionable practices, unlike MTRAs. So it is not only the epistemic circularity of an argument that makes MTRAs equally available for objectionable practices.
But what else is it that makes them equally available? Again, mentioning the fact that MTRAs fail to exhibit a doxastic practice's self-support is not needed in order to explain their equal availability. Epistemic circularity alone does the job. And if it is epistemic circularity that explains why MTRAs are equally available for bad practices, then it is epistemic circularity that explains why MTRAs are (and ought to be) under a ban.
So if Alston holds that arguments that exhibit a practice's self-support are acceptable, they are thus solely in virtue of their self-support, or so I think Alston is committed to holding. This is an important point to realize.
Note now that in order to make it clear that SP does, in fact, display significant self-support, one must make use of sense-perception; this is why such arguments for the reliability of SP are epistemically circular. But if Alston uses self-support as a way to distinguish virtuous from vicious epistemic circularity, then it is the fact that the argument does exhibit self-support that neutralizes any erstwhile vicious epistemic circularity. And in that case, the very property (self-support) that is supposed to neutralize epistemic circularity itself makes the argument epistemically circular.
Suppose, then, that I maintain (on any grounds whatsoever) that every instance of epistemic circularity is vicious. Then I shall not be willing to take Alston's proffered method of distinguishing good from bad circular arguments seriously at all. Perhaps if Alston had a way to distinguish the virtuous circular arguments that did not involve committing circularity, I might be persuaded. In short, Alston's way of distinguishing virtuous and vicious epistemic circularity is itself question-begging.
This point can be brought out in a slightly different way, not by assuming that all instances of epistemic circularity are vicious, but by assuming that they're all virtuous. Suppose I am generous I want to welcome all doxastic practices into the fold of the intellectually respectable. (So I maintain that tea leaf reading, crystal ball gazing, wild guessing, and so forth are all perfectly reliable.) Fortunately for me, I can produce arguments for my views, but they are epistemically circular track record arguments. Next I learn that, according to Alston, my MTRAs are under a "ban" precisely because such arguments are equally available for the practices that, it so happens, I maintain are reliable. I demand to know why Alston has the ban in place. I am told that those practices do not display significant self-support, while SP, for example, does. But I am not persuaded. I demand to know why it should matter that a practice doesn't display self-support, i.e., why a circular argument that doesn't exhibit the practice's self-support is, hence, viciously circular. How can Alston reply?
Alston says that SP displays significant self-support relying on SP permits a high degree of reliable prediction and control while crystal ball-gazing does not.  But so what? Alston cannot distinguish virtuous from vicious circularity by his proposed method without assuming that some practices are, indeed, correctly regarded as disreputable. I contend that Alston simply begs the question in the favor of the reputable practices. After all, what makes his criterion for distinguishing viciously from virtuously circular arguments plausible is that arguments supporting practices we believe, antecedently, to be illegitimate are ruled unacceptable.
If Alston's point were that circular track record arguments for reliable practices are acceptable simply because the practices are, in fact, reliable, then he might have a way to draw the distinction. This might be what Alston's point reduces to, if the above line of argument is correct.
Interestingly enough however that might be this is just how Ernest Sosa does draw the distinction:
Why not distinguish the [crystal ball] gazers and the [ordinary] perceivers in that, though both reason properly and attain thereby coherence and justification, only the perceivers' beliefs are epistemically apt and constitute knowledge? On this view, the crystal gazers differ from the perceivers in that gazing is not reliable where perceiving is.... [T]he perceivers can know their theory to be right when they know it in large part through perception, since their theory is right and perception can thus serve as a source of knowledge. The gazers, being unreliable, cannot serve as a source of knowledge.
Epistemic circularity is acceptable, Sosa boldly proposes, but only when the process described in the conclusion really is reliable. The conclusion might have to be true in order for the premises to be true, but this doesn't matter because the conclusion is true. That is not the case with crystal ball-gazers. So that is how an otherwise upstanding epistemically circular argument can avoid the unfavorable comparison with an argument about crystal ball-gazing.
The correct response to this is obvious namely, that Sosa begs the question. Exactly how he begs the question is worth making explicit. Sosa implies in his article that he can infer from the claim that perception is reliable (call this claim "PR"), the claim that an epistemically circular argument for the reliability of perception is not viciously circular (call this claim "NVC"). Without even glancing at the details of the inference from PR to NVC, we can see immediately that PR requires support (by Sosa's own lights). And of course, it is an epistemically circular argument that supports PR (again, by Sosa's own lights). Sosa can thus be justified in believing that PR only if this epistemically circular argument for the reliability of perception is not viciously circular which is to say, Sosa can be justified in believing that PR only if NVC is true.
Hence, Sosa's argument that a particular inference is not viciously circular is itself a textbook example of epistemic circularity. This is clever, perhaps, but unacceptable. We could grant that it might be the case that some epistemic circularity is benign. It will not do, in any case, to try to resolve that issue with an epistemically circular argument; any philosopher with his wits about him will reject any philosophical argument that depends on its own conclusion.
Sosa is expecting this sort of reply. As he says, "According to Barry Stroud, the perceivers can at best reach a position where they can affirm the conditional proposition that if their perception is reliable, then they know." Sosa is unimpressed; he is perfectly willing to assert the antecedent: "Perception is, of course, reliable while gazing is not. Therefore, the perceivers are right and apt both in their particular perceptual beliefs, at least generally, and in their theory of knowledge for it all rests in large measure on their reliable perception."
But Sosa's way of distinguishing benign from vicious circularity is successful only if epistemic circularity is not always vicious. This is another conditional proposition. Analogously, Sosa might make the smug assumption that epistemic circularity is not always vicious. But this is not as easily or uncontroversially assumed as that perception is reliable. So the trouble now is that his argument will not convince anyone such as myself who is not willing simply to assume that epistemic circularity is not always vicious.
Moreover, for anyone such as myself armed with positive arguments that epistemic circularity is always vicious, Sosa's way of distinguishing benign from vicious circularity will appear to be simply unjustified, since the premise, PR, is not justified by the (epistemically circular ) argument offered for it.
To sum up. Suppose that epistemic circularity is benign and that the mere fact of its circularity does not disqualify it as a candidate for justifying beliefs. Then we may construct epistemically circular track record arguments for standards regarding crystal ball-gazing, tea leaf -reading, and similar dubious belief-forming practices, according to which such practices could be responsible for justified beliefs. If tea leaf -reading results in justified beliefs, then Helga the Mystical is justified in the premises she forms based on the use of tea leaf -reading; and then her track record of (unsurprisingly) successful tea leaf -reading can justify her belief that tea leaf -reading gives her justified beliefs. Now, if epistemic circularity is benign, then nothing is wrong with Helga's argument as far as its epistemic circularity goes. But that's absurd; surely there is something wrong with Helga's argument, and it certainly appears to be due to the fact that it begs the question in a peculiar way. Hence epistemic circularity is vicious. Nothing Alston or Sosa has said successfully controverts this argument.
VI. The Arbitrary Case Argument.
My next, third argument that epistemic circularity is vicious follows an unusual strategy, and it should be helpful to lay out this strategy in advance. I stipulate that an argument is epistemically circular and I make some observations about that argument. Then I explain that if my observations about this first argument are correct, then it follows that a second argument is constructible. Then I stipulate further, for reductio, that the first argument's epistemic circularity is benign, from which it follows that the second argument is cogent; but this second argument plainly is not cogent. Hence, I conclude, the stipulation that the first argument's epistemic circularity is benign is false. But the first argument was an arbitrarily chosen argument. Hence epistemic circularity in general is vicious.
Here we go then. Suppose we have a track record argument A for some j-standard, and let us stipulate that the argument is epistemically circular. A contains this pair of premises:
(1a) S's belief that p1 meets conditions c
at time t1.
(1b) S is justified in believing that p1 at t1.
The conclusion is:
(C) If S's belief that pn meets conditions c at time tn, then S is justified in believing that pn at tn.
For someone advancing A as that which justifies his belief, his belief in (C) is justified only if beliefs in (1a) and in (1b) are both justified. Bearing in mind that A is epistemically circular, we may say that the truth of (C) is a necessary condition of S being justified in believing at least one of the premises or that the inference is valid. So let us suppose that the truth of (C) is a necessary condition for the claim that (1) (that two-part premise in particular) is true. Given this supposition, it follows from the claim, "S is justified in believing (1)" that the conclusion is true.
A reductio ad absurdum can be constructed here. In addition to the above stipulations, suppose further that epistemic circularity is benign. In that case, we may say that S has justified his belief that (C) on the basis of beliefs about A. Hence S is justified in believing all of A's premises. Moreover, recall that (1) is a premise such that S's being justified in believing it implies the truth of the conclusion.
If these stipulations are correct, then we are committed to the cogency of another argument, as follows: we state the premise "S is justified in believing (1)," from which we immediately deduce that (C). But this argument is obviously unacceptable. (C) does not follow from "S is justified in believing (1)." Hence we may conclude the reductio by rejecting the assumption that epistemic circularity is benign. It is vicious; saying so allows us to avoid the claim that S's being justified in believing (1) does imply, by itself, that (C).
It should be obvious that I am not claiming that the track record argument, as introduced by S, is itself a deductive argument. I am claiming, instead, that one needs only one pair of premises such as is found among the premises of a track record argument, and the assumption that epistemic circularity is not vicious, and the conclusion does follow deductively.
The point bears restating in other words. Suppose again that A is epistemically circular and that the circularity is benign. Suppose this benignity allows S to justify his belief that (C) on the basis of beliefs about A. Hence S is justified in believing (1) (arbitrarily chosen). Therefore, probably, (1) is true. But (1) was arbitrarily chosen. Hence, by universal introduction, together with the fact that the original argument was benignly circular, we may deduce that (C) immediately. But that is simply too fast. In drawing this conclusion, we have not used any other of the premises; we have simply assumed that S can have a justified belief on the basis of the original (circular) argument.
And this suggests that the original track record argument was viciously circular. Surely the j-standard does not follow from one arbitrary pair of premises. So if the mere fact that a track record argument's epistemic circularity is benign implies that an argument can be constructed where such an inference can be made, then epistemic circularity is vicious.
To bring the point out in a rather different way, consider the following thought experiment. Suppose we are given data from some psychologists who are testing the reliability of a queer belief-forming process, QP. The psychologists tell us that their subjects use QP, but we have no idea as to what the process is. We can ourselves check as to whether the beliefs are true (we are told that we are not using the tested process in doing so). So we can formulate premise pairs to the effect that QP was used in forming the belief that p at t (as the psychologists tell us), and that p is the case (as we confirm for ourselves); similarly for q, r, s, etc. Now, we shall not be the least bit tempted, after one instance of this sort of premise pair, to conclude hastily that QP is reliable. And indeed, we would be no less tempted to conclude that QP is reliable if we knew that the experimental subjects were justified in their beliefs about one case. If we were to conclude this, then we would be seeming to beg the question very objectionably.
Compare this to the case of an obviously epistemically circular track record argument, for the reliability of SP, say. We know right away from the fact that it is epistemically circular that it just does not matter how many cases of sense-perception that we consider; if we can in even one case justifiably assert both that it perceptually appears, in certain circumstances, that p, and that the belief that p is justified, then we (philosophers, who see the epistemic circularity ) can with just as much or little justification assert that the conclusion is true. The particular premise pair that one adduces is arbitrary, because the pair will be confirmed in every case: the methods we use to determine truth are similar in every case.
In brief: the view that epistemic circularity is benign allows that an epistemically circular argument's conclusion follows very quickly, when we know that if it follows at all, it does not follow so quickly.
VII. The Supports Argument.
In my final argument, I shall try to show that epistemic circularity makes it impossible for anyone to construct a good argument for certain j-standards or for the reliability of the associated processes. So if beliefs in those j-standards are justified, they are not justified by beliefs about epistemically circular arguments.
There are a number of constraints that one might reasonably expect successful theories of justification to fulfill. One such constraint what might be called the supports constraint is that the alleged justifiers for a belief (whatever sort of thing they might be) support the belief they justify. I take it that this is uncontroversial.
Consider then what I believe is a consequence of the supports constraint:
(SC1) It should be possible in principle to construct cogent arguments for the (at least) probable truth of justified beliefs by adverting, in the premises, to the alleged justifiers of those beliefs.
This is simply a necessary condition of a justified belief having justificatory support. If such arguments are not constructible, the theory of justification in question is to be rejected. So, even on an externalist theory of justification, it should be possible to construct arguments of the sort described. This is not to say that anyone has to have access to such arguments, or that the premises of such arguments are beliefs in anyone's doxastic system; it is only to say that the theory, externalist or not, must permit that such arguments be constructible. Indeed, I think that if such an argument, for the probable truth of a justified belief, is not constructible (by God, an "ideal observer," or other suitable fiction) given the theory's resources, then in fact the theory fails to account for the justification of that belief.
Beyond that, however, I think that a proper formulation of the supports constraint will have it that the supports relation is nonreflexive: the alleged justifiers for a belief give nonreflexive support to the belief they justify. A belief cannot support itself. Correspondingly, we may restate the foregoing consequence of the supports constraint as follows:
(SC2) It should be possible in principle to construct cogent arguments that are not premise circular for the (at least) probable truth of justified beliefs, by adverting in the premises to the alleged justifiers of those beliefs.
Any justifier of a belief, whether or not it itself is an item that is believed, must be expressible in such a way that it gives noncircular support of the belief contents; I think that follows from the very notion of support.
But someone might wonder how (SC2) would be applied, for example, to a foundationalist theory. Consider basic perceptual beliefs, which, the foundationalist alleges, are directly justified not by beliefs but by experiences. In that case, when our ideal observer (not the person whose basic beliefs are now in question) "adverts" to the alleged justifiers of the basic perceptual beliefs in accounting for those beliefs' justification, whereof exactly does that adverting consist?
It does not consist of expressing the contents of a belief, which perhaps would be what adverting would consist of in the ordinary case. Rather, it might consist of describing the contents of an experience; that description then would bear some sort of logical relation of support to the contents of the basic belief. Exactly how our ideal observer might describe a raw experience we shall have to leave up to phenomenologists and logical positivists. But regardless of the solution, however indeed we make sense of various sorts of item supporting a belief, surely our notion of support is a logical notion, i.e., essentially it amounts to an argument being constructible between a description of the supporting item and the supported belief.
Now let us consider how these constraints, together with the problem of epistemic circularity, wreak havoc on an epistemically circular argument for a sample theory of justification: reliabilism. Let us suppose that the conclusion, a reliabilist j-standard, is a claim of the general form:
(RP) If S's belief that pn is a result of process R at time tn, then S is justified in believing that pn at tn.
Consider the premise pair (1a) and (1b) again but now suitably modified to support (RP). What lends support to the fact that (1b), viz., that S's belief that p1 is justified, according to our reliabilist? It is the fact that the belief is the result of process R.
It follows that S has a belief in (RP) that is based on an argument with a premise viz., (1b) that is believed justifiably in virtue of R's being reliable. But then we apply (SC2) to the situation. It looks as though our reliabilist is committed to the view that the content of the belief in the reliability of R i.e., belief in (RP) is inferentially supportable by a claim that R is reliable (i.e., what justifies S's belief that p1). That is false; the supports relation is not reflexive.
But we are forced to this conclusion whenever we accept that epistemic circularity is benign and offer an epistemically circular argument in support of our favored j-standards. If epistemic circularity is benign, then I can succeed in getting a justified belief from an epistemically circular track record argument; and what successfully justifies me in believing one or more of the premises is the conclusion of the argument. But that implies that the content of the conclusion can support itself (with the premises as intermediaries). And that is simply not how the supports relation works.
Hence, epistemic circularity is vicious. To suppose epistemic circularity to be benign would be to commit to something false, viz., that the supports relation can be reflexive.
Let us put this chapter's work into the context of the Meta-Regress Argument.
One way to end a regress of j-standards is to suppose that we can get justified beliefs in some standards via arguments that are licensed by those same standards. Such standards would be, in the jargon introduced in Chapter 1, self-supporting. The view that self-supporting j-standards exist is called `self-support meta-foundationalism' or `self-supportism' for short. As explained in Chapter 1, the arguments in virtue of which given j-standards are considered self-supporting are, necessarily, epistemically circular. Hence, if it should happen that epistemic circularity is vicious, i.e., that one cannot get a justified belief in a j-standard via an epistemically circular argument, then no j-standard can be successfully self-supporting. The aim of the present chapter has been to argue that indeed epistemic circularity as such (as opposed to only varieties thereof) is vicious. So if this chapter's arguments have been adequate to their aim, self-supportism is false.
It should be useful to reiterate why self-supportism is so plausible to so many. We can and do use basic belief-forming practices in confirming all sorts of beliefs. Indeed, it is overwhelmingly plausible to think that we are limited, in forming our rational, warranted, or justified beliefs, to a fairly small set of such practices (if they are construed generally enough). So why not think that those same practices can be applied to themselves, particularly because they work very well everywhere else? And then why not think that some basic j-standards could not, by the same token, support themselves?
This chapter's arguments should have made my answers to these questions clear. I will add now, however, that to suppose that basic belief-forming practices cannot confirm their own reliability is not to impugn the success of basic belief-forming practices in general. There is nothing contradictory or even strange, I would suggest, about the notion that a given practice cannot confirm its own reliability, or that basic j-standards cannot support themselves. Common sense and intuition have nothing directly to say on these issues. So we might as well follow the best arguments on these issues and thus, I maintain, reject self-supportism.