JONATHAN L. KVANVIO
CONSERVATISM AND ITS VIRTUES
1. INTRODUCTION
Several recent works have claimed that epistemologists ought to embrace some sort of conservatism doctrine with regard to beliefs.1 The doctrine in its starkest and, I shall contend, most implausible form claims that a belief has some presumption of rationality simply because it is held. A more careful formulation of the doctrine is this: some degree of rationality can be generated for a person=s belief merely in virtue of the doxastic commitments of that person.
The sort of conservatism doctrine which I wish to discuss is one of interest to epistemologists. The interest of the epistemologist is in the nature of knowledge, and thereby in that sort of rationality which is a necessary condition for knowledge. In addition to more practical sorts of rationality, there is a purely epistemic type of rationality which is characteristic of knowledge. Thus, there is a corresponding type of purely epistemic conservatism: that mere doxastic commitment can create some degree of purely epistemic rationality for some beliefs.
We can become clearer regarding the nature of epistemic conservatism by contrasting it with other positions in the conceptual neighborhood. First, there are other sorts of conservatism doctrines. One view might be called practical, or pragmatic, conservatism. Such a position holds that it is the practical value of being conservative about our beliefs that creates the presumption in favor of the (practical or pragmatic) rationality of belief. There are many sorts of arguments that might be given for such a view: one might argue that it is a rather tiresome effort to keep altering beliefs from moment to moment, or that the survival of the species hinges on our being conservative, or that being conservative is more likely to get us to the truth in the long run.2
Another sort of conservatism doctrine is a purely descriptive one. Quine provides a good example of this use.3 On Quine=s view, there is no normative significance to the discipline of epistemology, and since there is not, one might as well settle for psychology. Since it is a psychological fact about humans that they are slow to alter their beliefs, and that, when choosing between alternative hypotheses, the one that requires the least amount of alteration in the rest of one=s belief system is preferred, it follows that a purely descriptive conservatism doctrine is true.
Another conservative position from which epistemic conservatism differs is Davidson=s semantic conservatism. According to Davidson, it is given that most beliefs are true. His argument for this claim is that the content of a belief is determined by its location in a pattern of beliefs. If the pattern itself is assumed to include mostly false members, this assumption undermines the attempted specification of the subject matter of the original belief, and without a determinate subject matter, there is no belief to be either true or false.4
Davidson=s semantic conservatism differs from epistemic conservatism in that, whereas semantic conservatism holds that any belief is presumptively true, epistemic conservatism holds that (some) beliefs are presumptively rational. According to epistemic conservatism, mere doxastic commitment creates some degree of epistemic rationality; semantic conservatism holds that for a pattern of beliefs to include false members, most of the rest of the pattern must be true. Now surely there is some sort of connection between what is epistemically rational and what is true, but the connection is not so straightforward that one can identify what is true, or even most likely true, with what is epistemically rational to belief.5
There are interesting connections between semantic and epistemic conservatism, nonetheless, for if both views could be defended, we would have the beginnings of a strong argument against universal skepticism. For if it could be defended that most of our beliefs are most likely true and at least presumptively rational, we are well on our way to showing that most (or at least some) of our beliefs are instances of knowledge.
Another view which needs to be distinguished from epistemic conservatism is the view that epistemic justification (or rationality) is subjective. As I understand it, a subjective theory of justification holds that only features which are >in the head= of the appropriate person are relevant to the justificatory status of that person=s beliefs. This construal makes such a view look much like epistemic conservatism, for epistemic conservatism is clarified using the notion of mere doxastic commitment C and surely what one is doxastically committed to is something >in the head= of the relevant individual. Nonetheless, the two views are distinct. First, a subjective theory of justification can allow that there are other features to the notion of what is >in the head= other than what one is doxastically committed to; a subjectivist might, for example, count dispositions to believe in certain specified circumstances as relevant to justification. One sort of subjective theory that includes such conditions is a theory which implies that a belief is justified for a person if that person would believe that that belief is justified, were he to reflect on the justificatory status of that belief.
Epistemic conservatives cannot appeal to such dispositions to believe, for epistemic conservatism requires that the level of rationality in question accrue from actual doxastic commitments. Dispositions to believe are dispositions toward doxastic commitments, not actual doxastic commitments.
Another way in which conservatism and subjectivism differ concerns what sorts of theories of justification conservatism can be a part of, i.e., what additional conditions an epistemic conservative might cite in describing how a presumptively rational belief comes to be fully rational. According to subjectivism, only features regarding what is in the head of the individual are appropriate. Though a conservative might agree with the subjectivist on this point, he need not; for as I construe epistemic conservatism, a conservative is not barred from including any sorts of features in his complete theory of justification.
This last point is important to note since overlooking it can lead to the mistaken view that conservatism is wedded to an internalist, as opposed to an externalist, view regarding justification. Internalism is the view that justification is determined for a person on the basis of subjectively accessible conditions, externalism is the view that justification can obtain independently of a cognitive grasp of the elements which do the justifying. So construed, internalism specifis a class of theories of justification of which subjectivism is one subclass; and, just as conservatism is not wedded to subjectivism, it is not wedded to internalism either. The reason here is the same as above: conservatism does not specify in any way what additional components are to be added to turn a presumptively rational belief into a fully justified belief C no sort of internal or external condition is excluded by the doctrine of conservatism. I shall have more to say about the connections between conservatism and internalism later.
2. THE VIRTUES OF CONSERVATISM
Epistemic conservatism might seem initially quite implausible. After all, why should the mere fact of doxastic commitment result in any positive epistemic status for any belief? In spite of its prima facie implausibility, conservative viewpoints are presupposed by quite a broad range of epistemological positions. First, several standard accounts of philosophical methodology presuppose the truth of some sort of conservatism doctrine C in particular, maximalism, Rawls=s method of reflective equilibrium and Chisholmian particularism. Maximalism is the view that in formulating a theory, we are entitled to employ all of our antecedent beliefs in determining which methods to follow and which not to follow. Goldman says,
Maximalism invites us to use all our antecedent beliefs whenever we wish to appraise our cognitive methods. A maximalism argues there is likely to be little or no choice among methods unless we employ a prior corpus of beliefs. And if some prior beliefs are allowed, why not allow them all?6
Rawls=s method of reflective equilibrium is the view which claims that in constructing a theory, one begins with claims that seem most obviously true. Then one arrives at a theory by eliminating the inconsistencies, always being careful not to give up the more obvious in order to retain the less obvious.7 Chisholm=s particularist methodology begins with particular judgments of the form >belief B is justified in circumstances C=, and attempts to form generalizations which account for these initial judgments.8
A question which any proposed philosophical methodology must answer is this: how is it that the use of the proposed methodology is not simply arbitrary, i.e., how is the use of this methodology supposed to generate results which call for our rational allegiance? All sorts of ridiculous methodologies can be proposed (for example, those still longing for the days of yore may propose beginning with what was believed by their favorite forefather and making as few changes as possible); but how is any methodology supposed to generate results which are rational to believe? And, for the proposed methodologies above in particular, how is the use of all of our beliefs as a starting point supposed to generate results which are not the result of a purely arbitrary starting point? The doctrine of epistemic conservatism provides, as far as I can see, the only plausible answer for these positions. The reason that any of these starting points is not purely arbitrary is that the mere fact that one holds a set of beliefs creates a presumption in favor of the rationality of at least part of that set of beliefs. For, if those beliefs had no presumption in their favor, or if there was a presumption against those beliefs, it is very hard to see how the application of any of those strategies is to generate a final set of beliefs that are not simply as defenseless as the starting point itself.
Further, even if there is some other way to defend these methodologies from the arbitrariness charge, conservatism still offers the hope of a clear defense of these methodologies from that charge. So, even if some other defense of these methodologies is possible, embracing a conservatism doctrine would be advantageous for defenders of these methodologies.
Besides being methodologically presupposed, conservatism is also substantively presupposed by some natural and intuitive responses to skeptical arguments. For example, some claim that when confronting two incompatible but otherwise equally adequate hypotheses, an attractive position holds that one ought to accept the one that requires the least alteration in one=s present set of beliefs. For example, consider Morris=s account of why Russell=s hypothesis that the world is only five minutes old is one that we rationally reject:
So what is wrong with Russell=s hypothesis? Accepting the truth of that proposition would require rejecting the truth of the vast majority of all the empirical beliefs we hold.... In general, the more of our present beliefs, and especially our fundamental beliefs, an hypothesis accords with or accounts for, the more reasonable it is for us to accept. The fewer of our beliefs an hypothesis preserves, the less reasonable it is to accept. The five minute hypothesis accords with only a minute fraction of our factual beliefs, and oilers absolutely no account of why we hold the others. So although the hypothesis cannot strictly be proved false, it is such that it would not be rational for any normal person to accept it.9
Again, it seems that such a view presupposes an epistemic conservatism doctrine, otherwise one would think that each of the alternative hypotheses would be equally rational to affirm. What makes the hypothesis more rational which accords with or accounts for (more of) our present beliefs, it would seem, is simply that our present beliefs are at least presumptively rational. If they were not, it is not clear why preserving them would be rational. Even if this view does not presuppose conservatism, a conservatism doctrine clearly explains why such a procedure is at least presumptively rational: simply because one has the beliefs in question makes retaining them presumptively rational.
Another way in which conservatism is presupposed is in a common reaction to skeptical arguments, like Russell=s above, which share what I shall call the alternative hypothesis thesis (the AHT). Very roughly, the AHT is the claim that, because there is a satisfactory alternative account to some of our ordinary beliefs, what we normally accept is no better off than the alternative view.
Quite often upon confronting such arguments we cannot discover a defect with the alternative hypothesis, and yet we go on believing what we presently believe anyway. Some may argue that this is merely a result of our lack of direct control over our beliefs which brings us as creatures of habit into the unhappy situation of holding ungrounded beliefs. A different account of the matter is that it takes more than an equally adequate alternative hypothesis to undermine our present beliefs. If epistemic conservatism is true, the hypothesis that we accept can be epistemically superior to the alternative hypothesis proposed; and thus the argument employing the alternative hypothesis is a bad argument since it does not follow that our present views are no better off than the alternative simply because the alternative can equally well handle all of the data to be explained.
Further this implication of conservatism seems right. The existence of alternative hypotheses may rightly prompt the need for further research to determine why the one hypothesis is unacceptable, perhaps to avoid a justified charge of narrowmindedness. However, if we encounter such an alternative hypothesis which conflicts with what we normally accept, the rationality of our view is not necessarily affected by the mere existence of such a hypothesis. For example, in evaluating new philosophical views, we quite often believe them to be false when they conflict with what we already believe in advance of figuring out exactly why the view in question is false. Such a procedure need not be irrational. One might attempt an explanation of this fact without presupposing conservatism, but conservatism provides an attractive account of the matter. We can rationally reject a view because of its conflict with our present views even though we have nothing which we can point to at present to show that the view is false. There probably are those who succumb to Al-IT arguments and give up their view that their present views are shown to be true. But an at least equally plausible response is to maintain that one=s present view is superior even though one cannot at the moment pick out why. And if this latter belief is reasonable, it seems quite clear that it is more reasonable to reject the alternative hypothesis and stick with one=s original view.
It is also tempting to think that the doctrine of epistemic conservatism is presupposed by any plausible non-skeptical position in epistemology. For any alternative to skepticism which enjoys the image of a piecemeal revision of our raft of beliefs will presuppose that there is some presumption in favor of the parts of the raft not under revision, and the attempts at specifying the ground of this presumption in terms of some sort of coherence relation have not been especially successful. Perhaps the lesson here is that the only alternative to the infallibilism the raft metaphor meant to oppose involves the doctrine of conservatism: that what grounds the presumption in favor of the parts of our belief system not presently under revision is the doxastic commitments we have.
We can see then that epistemic conservatism is a quite important doctrine, possessing significant explanatory virtues. Nonetheless conservatism is not intuitively attractive and the sorts of defenses of it that have been offered have not done much to make the view more attractive. I wish to remedy this situation. To do this, I shall first show how problematic are the standard versions of the conservatism doctrine. Seeing the defects of these accounts will aid us in developing an adequate account, one for which a strong argument can be constructed. The particular conservatism doctrine I shall offer has implications for the earlier discussions of philosophical methodology and the rationality of rejecting alternative hypotheses, and I shall conclude with a brief discussion of these implications.
3. PURELY EPISTEMIC ACCOUNTS OF CONSERVATISM
Most versions of the conservatism doctrine are clearly implausible. One sort that is obviously inadequate is unrestricted conservatism. Unrestricted conservatism is that type of conservatism which does not restrict the class of beliefs in any way which are claimed to have some degree of positive epistemic status. One example of such a principle claims that, necessarily, any belief of a person is rational.
Unrestricted conservatism is clearly inadequate, though, for if it were true, it would be impossible to ever have an irrational belief. But that is quite counterintuitive, for most of us can recognize that at least sometimes we have held irrational beliefs. Thus, unrestricted conservatism fails.
The most explicit defense of conservatism in the literature is a version of restricted conservatism, and is made by Chisholm. His formulation of it is:
Cl: Necessarily, if S believes p and nothing else S believes explicitly contradicts p, then believing p has some presumption in its favor for S.10
Chisholm=s version of conservatism is an advance over unrestricted conservatism in one way, for it seems to allow that some beliefs can be irrational. For, any case where S has other beliefs which explicitly contradict his belief that p can be a case where believing p is irrational.
However, given Chisholm=s understanding of explicit contradiction, S will have other beliefs which explicitly contradict his belief that p only if S also believes not-p.11 Problems arise because of this understanding. First, contradictory beliefs are not possible, i.e., it is not possible that S both believe p and not-p at the same time.12 If I am right about this, then Cl has equivalent results to those of unrestricted conservatism, for there will no longer be any possible cases on Cl in which S irrationally believes that p. Second, and more important, even if contradictory beliefs are possible, there are cases in which a belief can be irrational even though there is nothing which explicitly contradicts that belief. As Foley claims, in discussing Chisholm=s view,
But even Chisholms principle is far too strong, as can be seen by considering a person whose beliefs are not marred by any explicit contradiction but who believes a proposition when given his circumstances it is more reasonable to believe its negation.... Yet this is just what Chisholm=s principle says is impossible .... But, this is strongly counterintuitive. Unfortunately, people do believe propositions whose negations are more reasonable for them to believe.13
There are numerous examples that could be constructed of the sort Foley discusses, but the following example will be helpful in showing where we must look for an adequate conservatism doctrine. Suppose that Joe believes that God exists while at the same time reasonably believing the truth (let us suppose) that there are overriding epistemic grounds against this belief. In reflecting on this situation, Joe comes to reasonably believe that his belief that God exists is an epistemically irrational belief. Now, Joe may believe, and it may actually be true, that there are non-epistemic grounds which nonetheless imply that his belief is rational on the whole (these facts may explain why he continues to hold the belief). For example, in Pascalian fashion, it may
be more reasonable given the benefits and risks associated with the existence of God to believe that He exists. Even if that is true, this much is clear: if one considers epistemic matters alone, Joe=s belief that God exists is not epistemically more reasonable than believing that God does not exist. For, if we discount any possible non-epistemic rationality-conferring properties in this case, Joe would be quite unreasonable in failing to hold that God does not exist.
The way in which this last example undermines Cl shows that no purely logical relation between beliefs can salvage the conservatism doctrine. For any logical relation between beliefs placed in the antecedent of Cl to circumscribe which beliefs are to be counted as presumptively rational, that antecedent will still be satisfied by Joe=s beliefs that God exists and that there is sufficient epistemic reason to believe that He does not. Neither entails the negation of the other, nor are they even inconsistent with each other. We might think that probability connections are the sort to be looked for, and it may be that there are probabilistic connections between seeming-states and claims about the external world. However, the difficulties with interpretations of probability theory are too well-known to allow any such probabilistic connections to be of such use in defending conservatism. For example, the interpretation that I find most satisfactory C the empirical interpretation C has the flaw of making probability connections a contingent matter. So, on this interpretation, whether seeming-states make probable claims about the external world depends on what sort of a world we are in (in Descartes= evil demon world, for example, no such probability connections obtain). Conservatism, if true however, is not contingently true; it is rather a part of the essence of rationality that mere doxastic commitment implies some level of rational acceptability. Thus, probabilistic connections will not be of much use to the defenders of conservatism even if such connections obtain.
We might try at this point to insert a purely negative requirement in the antecedent of our conservatism principle. The weakest such principle we might try is:
C2: Necessarily, if S believes that p and does not believe that his believing p is epistemically suspect (i.e., has negative epistemic status), then believing p has some presumption in its favor for S.
C2 is an attempt of the sort we must look for, for the restrictive clause in the antecedent is not of a logical or probabilistic sort, and it also avoids the counterexample of Cl. For in that case, Joe believes that his belief that God exists is epistemically irrational, and hence he believes that his belief is epistemically suspect. Thus, the antecedent of C2 is violated, and it is thereby not a counterexample to C2.
However, C2 will not do, as the following case shows. Suppose Jack, upon having a near-death experience, finds himself believing that disaster will strike the earth in 1999. He is absolutely convinced that this is so, but grants upon inquiry that he has no reason whatsoever for thinking that it is true. ACall me what you will@, he says, Aa fideist, irrational, loony, or whatever; but I am as certain that such a disaster will occur as anything, even though I realize I have no reason to think so@.
Should we say, as C2 implies, that Jack=s belief has some presumption in its favor for Jack? I should think not; in fact, even Jack himself recognizes that there is no such presumption. Since C2 implies that there is such a presumption, we must reject C2.
Our rejection of C2 is not the end of our search for an adequate conservatism doctrine; though, I contend, it is the end of the road for what we might call first-order conservatism. A first-order conservatism doctrine is one that claims that the presumptive rationality that accrues in a fashion in accord with conservatism, accrues on the basis of first-order beliefs C beliefs which are not beliefs about beliefs (second-order beliefs), or beliefs about beliefs about beliefs (thirdorder beliefs), etc. No objective restriction between first-order beliefs is successful in specifying when presumptive rationality accrues, and the attempt of C2 to maintain first-order conservatism by adding the absence of certain second-order beliefs fails as well. If there are other ways of defending first-order conservatism, it is quite difficult to imagine what they might be. I take it then as fairly well established that first-order conservatism fails.
The situation is not hopeless though, for epistemic conservatism is not necessarily a first-order view. My suggestion is that we attempt to save epistemic conservatism by offering a version of it which is not a first-order version. One such version is:
C3: Necessarily, if S believes that p and believes that S=s belief that p is shown to be true, then S=s belief that p has some presumption in its favor for S.
Even though C3 is not a first-order version of epistemic conservatism (since its antecedent includes a second-order belief), it is still a version of epistemic conservatism. As I have clarified the doctrine, epistemic conservatism claims that the presumptive rationality of a belief can arise from mere doxastic commitment. Normally, this claim is interpreted so that the doxastic commitment is identified with the belief in question, but I see no reason to require this identification: epistemic conservatism is the view that the phoenix of rationality can arise from the ashes of mere belief (without the additional constraints that the beliefs have a content which provides evidence, or some sort of epistemic support or ground for the fledgling rationality). If we do not make the mistake of identifying the doxastic commitments in question with the particular belief that is supposed to be presumptively rational, we can see that C3 is a version of epistemic conservatism in that it specifies certain doxastic commitments (those in the antecedent) which necessarily result in the presumptive rationality of a certain belief.
I propose that an acceptable version of epistemic conservatism is to be found in an investigation of the sort of account 9resented in C3, what we might call higher-order conservatism (C3 itself would be an instance of second-order conservatism). Both types of conservatism share the claim that rationality can arise merely through forming and holding beliefs; a distinctive claim of higher-order conservatism is that rationality arises through thinking, or having thoughts, about the epistemic status of one=s beliefs. Higher-order conservatism implies that rationality can, in a sense, be created ex nihilo: beliefs can come to have a degree of rationality merely by being thought about even though, in themselves, those beliefs possess no epistemic merit at all, i.e., they are neither supported by any evidence nor are they rational apart from evidence by being self-evident, or intrinsically warranted in some sense (as are foundational beliefs). In the next section, I wish to present an argument for accepting higher-order conservatism and to consider whether C3 itself is the formulation of higher-order conservatism we should accept.
4. HIGHER-ORDER CONSERVATISM
The argument which I wish to develop attempts to show that if we are fallibilists about justification, i.e., if we think it is possible in general to have justified false beliefs, then we ought to embrace higher-order conservatism. We have already seen that no other sort of conservatism doctrine is plausible; what is left, then, is to establish the connection between fallibilism and conservatism.
Those cases in which conservatism seems the most implausible are cases in which the person has improperly handled his evidence. Perhaps he ignores some, perhaps he weighs it poorly. In any case, the person in question is mistaken about what the evidence really confirms, and in particular, his belief runs contrary to what the evidence confirms. To some it may seem that in such a case, even if the conditions in the antecedent of C3 are satisfied, there is no presumptive rationality for such a person regarding the belief in question.
There are many cases of this sort that need not detain us long. In particular, it should be noted that a person can clearly have a rational belief if the evidence in question is not avilable to him. It is only evidence which we are epistemically responsible for having that affects the rationality of our beliefs. Further, it is not the mere possession of evidence that is critical, for one can, for example, watch an experiment designed to test a certain theory and yet not be privy to enough information to know what the result shows. In such a case, merely having the evidence (i.e., the outcome of the experiment) does not affect the rationality of one=s beliefs about the theory in question, one must also have enough information to be capable of rationally basing one=s belief about the issue in question on that evidence. This condition is clearly violated when one watches such an experiment, knowing it is relevant to a particular theory, but having no idea what any particular result shows about the theory.
Nonetheless, there are many cases that remain. There are persons who base beliefs on things obviously irrelevant to their beliefs C for example, an imaginative eight-year-old believing there is a goblin in his room because he just heard the door creak. There are persons who jump to conclusions before carefully evaluating all the evidence. There are persons who carefully evaluate all the evidence, but make bad inferences from the evidence. In all these cases, many will be tempted to think that there cannot be even the slightest presumption of rationality for such beliefs.
My contention regarding such a conclusion is that it is justified only if a certain aspect of infallibilism is defensible. To see this, considerthe sorts of discoveries psychologists make about the way in which we naturally draw inferences. We naturally generalize hastily, for example. In spite of using a bad (non-truth-conducive) inference pattern, I contend that we are nonetheless (sometimes) rational in drawing conclusions on the basis of hasty generalizations. Now, one might agree that there is some truth in this position and yet try to explain this rationality as practical rationality rather than epistemic rationality: one might claim that any rationality present accrues because the risks and benefits of not generalizing quickly outweigh those that arise from the fact that generalizing so quickly generates many false beliefs. That explanation ignores the complexity surrounding hasty generalization. What is involved in faulty generalizations is, first, not paying close enough attention to the degree of arbitrariness involved in the particular instance(s) from which one generalizes and, second, not following the statistical regulations on the degree of confidence one can place in a Sample class of a given size. But we are not born properly discriminating when certain arbitrariness criteria are satisfied and when they are not C we learn these when we begin to look for an explanation for why some of our procedures seem to be generating quite a few false beliefs C nor are we born with the knowledge of probability theory necessary to determine the confidence level appropriate to a given sample size. Until we become aware that such criteria are important for finding the truth and avoiding error, why should the fact that such criteria are important in this way affect the rationality of our beliefs? The mere fact that evidence exists that shows that my belief is false is irrelevant to the rationality of that belief unless one is aware of the evidence and sees its relevance to one=s belief; and just so with rules of inference C the mere fact that a rule of inference is not a reliable guide to finding the truth and avoiding error does not by itself imply that any belief arrived at through the use of that rule is not rational.
The opposite view, that using inappropriate inference patterns infects the rationality of the belief, is a holdover from our infallibilist epistemological past. We know, with regard to many inference patterns, which are good ones for getting to the truth and which are not. If we are infallibilists, we will hold that no belief is justified unless guaranteed to be true by the evidence (or by the self-warranting nature of the proposition or belief). This view about justification requires that an inference pattern cannot be employed to confer justification unless it is a good one, i.e., one successful in getting us to the truth. This assumption then is translated into a requirement that beliefs with any degree of rationality not be the result of following bad inference patterns, for, we would assume, no sort of justification can result from following a bad inference pattern. What follows, then, from these assumptions is that no degree of rationality can be present when one misweighs the evidence, believes contrary to the evidence, etc.; for, obviously, such cases involve the case of bad inference patterns.
We have come to see, though, that in general it is possible to have justified and yet false beliefs; we have come to see the truth in fallibilism. The instructive implication of fallibilism is that the links between truth and justification are at best indirect; nothing follows about the justificatory status of a belief merely given that it is false. But, if we can have justified false beliefs, why not about the rules of inference we employ? If our beliefs about inference patterns are not infallible (and they surely are not), what possible reason could be given for insisting that unless inferences are correct, they and the results of following them cannot be justified? Such a position is caught between a rock and a hard place: it is known that infallibilism is false, and yet it is insisted that an unwarranted implication of infallibilism be maintained.
Such a view maintains that unless a rule of inference is a correct one, no use of it, even if one justifiably takes the rule to be correct, can render its products justified. Consider the analogous point about evidence. Such a position would hold that even if a set of propositions or beliefs bears a justifying relation to a belief, that set cannot justify the belief unless every member of the set is true. False information within such a set bars that set from functioning as evidence for the belief in question.
My question is why anyone would be attracted to this position regarding evidence. The only remotely plausible answer commits such a position to an aspect of infallibilism. According to infallibilism, justification and truth are forever wedded. Hence, if one=s rules of inference are good ones, starting from false premises is bound to generate false beliefs. The proper lesson, according to infallibilists, would be that no set of propositions or beliefs can be evidence for a further proposition or belief unless the original set contains no falsehoods.
This view about evidence is quite implausible. Its implausibility is seen by noting that the only plausible defense of it appeals to infallibilism (which I am assuming to be false without argument). Yet, if this view about evidence is implausible because it depends on infallibilism, so is the analogous position about rules of inference. The only defense for requiring correct rules of inference in justification-conferring situations appeals to the bond infallibilists claim between justification and truth. In the absence of a commitment to infallibilism, no such restriction on rules of inference is defensible.
One might object here that there is another view regarding justification that can ground the required connection between justification and truth required for the above view regarding rules of inference. The view I have in mind is reliabilism. According to reliabilism, a belief is justified (roughly) if and only if it is produced by a mechanism which produces a sufficiently high percentage of true beliefs. If this view is true, it might be objected, don=t we have just the connection between justification and truth required for insisting that rules of inference be correct in order to confer justification?
Yes, if the view were true; but it is not, as can be seen by considering the following objection. Suppose our world is run by an evil demon, as gifted in power as he is intent on getting us to believe what is false. Even if this were true, we would still justifiably believe that there are trees, mountains, etc. The demon could make most of our beliefs false, but he would not thereby make our beliefs unjustified.
The version of reliabilism above fails in that it implies exactly what infallibilism implies regarding the justificatory status of beliefs in evil demon worlds: that beliefs in that world cannot have the justificatory status as beliefs in a non-evil-demon-world. In this respect, unbridled reliabilism and infallibilism are identical twins. Reliabilists sometimes attempt to distance themselves from this aspect of infallibilism by maintaining that they are offering a theory of justification that applies to our world, not just any world. This response fails to appreciate the force of the evil demon world argument. That argument can be taken to be about some other possible world distinct from our world, but it can also be taken to be about what our world might be like. When the evil demon objection is taken in that way, the unbridled reliabilism above is committed to exactly the same regrettable denial of justification regarding the evil demon world as is infallibilism. It is in precisely this respect that the only plausible defense of requiring true
inference principles is to be found. A proper rejection of infallibilism eliminates this possible ground of the position in question.
Note that in defending reliabilism against just the evil demon world objection, Goldman alters the rough theory above so that the connection it establishes between justification and truth is lost. Goldman says,
We have a
large set of common beliefs about the actual world: general beliefs about the
sorts of objects, events, and changes that occur in it. We have beliefs about
the kinds of things that, realistically, do and can happen. Our beliefs on this
score generate what I shall call the set of normal worlds. These are
worlds consistent with our general beliefs about the actual world. ...
Our concept of justification is constructed against the backdrop of such a set
of normal worlds. My proposal is that, according to our ordinary conception of
justifiedness, a rule system is right in any world W just in case it has a
sufficiently high truth ratio in normal worlds.14
The important point to note is that Goldman here replaces an emphasis on high truth ratio with an emphasis on high truth ratio in normal worlds. Further, the concept of a normal world is a subjective one: it is a world consistent with our general beliefs about the actual world. It is important for Goldman to alter the reliabilist theory in this way in order to avoid the evil demon objection above. In doing so, the connection reliabilism originally maintained between truth and justification is lost. A justified belief no longer need be one that is likely to be true.
Yet, if there is no general connection between justification and truth, there is no justification for insisting that rules of inference be true in order to convey justification. Goldman=s reliabilist theory of justification is the most developed of the various versions of reliabilist theories of justification, and it lacks this implication that a rule of inference must be true to convey justification. Hence, the point made earlier holds: only a regrettable commitment to infallibilism yields the views required to cast doubt on conservatism.
The lesson here is that cases of at least presumptively rational belief can arise both when we have evidence that person lacks and when we are better at drawing inferences than that person is. We have no immunity from error about what really confirms what any more than we do about the rest of our world in general. Confirmation relations are just as clearly something we learn about, something we can be mistaken about, and something we make progress in understanding as are other matters of fact. And this is the important point: this learning process can be infected to the core with rationality, so that the process itself can issue in products which are at least presumptively rational even before we arrive at the point where we see all things clearly.
How does this rejection of infallibilism help in tying conservatism to fallibilism? Consider again the problem cases for higher-order conservatism: persons holding beliefs on the basis of events completely irrelevant to their beliefs, persons weighing evidence improperly or simply making a mistake in evaluating what the evidence shows, and persons being hasty in drawing conclusions. First, if we are fallibilists, the mere fact that a person makes a mistake or is hasty in drawing in conclusion should not lead us to assume that the belief has no presumption in its favor for that person. What matters is, rather, whether, from his perspective such practices are epistemically suspect; and if he believes that his manner of arriving at a belief is suspect, it is hard to see how he could believe that his belief is shown to be true. Further, the same point holds for cases in which a person holds a belief on the basis of events clearly irrelevant to the truth of the belief. The critical question is: Aclearly irrelevant@ from whose perspective? If from ours, then we are illegitimately evaluating the rationality of others on the basis of information we have; if from theirs, then how is it that they also fail to think of their belief as epistemically suspect? In all these cases, it would seem that shared information is unwarrantedly presupposed in finding conservatism problematic; in the absence of that assumption, such cases show no problem for epistemic conservatism.
This unwarranted presupposition takes us back to infallibilism. The only plausible reasons for presupposing shared information are either that the information in question is known a priori by everyone, or that some feature regarding the nature of justification requires that such information be presupposed. The feature in question would have to be the connection seen between justification and truth. Neither reason justifies assuming shared information regarding rules of inference. Such rules are not known a priori by everyone, and the connection between justification and truth is not strong enough to support the presupposition. The only justification available for the required bond between justification and truth is a defective view concerning the relation between justification and truth inherited from infallibilism. Since that position on the connection between justification and truth is false, there is no justification available for this position regarding rules of inference.
So, a full break with infallibilism provides the needed justification for conservatism. Is C3 the version of higher-order conservatism we should embrace? It is not, for it is possible for a person to irrationally hold the second-order belief specified in the antecedent of C3. In that case a person believes a proposition, and irrationally believes that there is something that shows his belief to be true. In such a case, we should not hold, as C3 implies, that there is even the slighest presumption of rationality in favor of the belief in question. Hence, C3 is defective.
This problem is not difficult to remedy, though. Part of what I argued in tying conservatism to fallibilism is that it is beliefs or awarenesses (explicit and implicit) which determine at least the presumptive rationality of belief. So, if a person is to irrationally believe that there is something which shows his belief to be true, there must be something of which that person is aware (at least implicitly) which he takes (at least implicitly) to show that this belief is not a good one to have. So, we can handle the difficulty by replacing C3 with:
C4: Necessarily, if S believes that p, believes that there is something that shows that S=s belief that p is true, and believes nothing else that he takes to show that it is not the case that there is something that shows that S=s belief that p is true, then S=s belief that p has some presumption in its favor for S.
One question remains about this account, and it concerns the notion of something showing that a proposition is true. I have purposely picked this ordinary though somewhat ambiguous locution in order to prevent the appearance that only professional epistemologists fall under the scope of C4. The idea here is that a person thinks something shows a proposition to be true just in case that person believes that that something, whatever it is, makes it obvious that that proposition is true. Clearly, this sort of belief is one that even quite small children can and do have.
5. CONCLUSION
I have defended that C4 is a proper account of conservatism. It is an account of epistemic conservatism in that it implies that a belief can come to be presumptively rational for a person (in the epistemic sense of >rational=) solely in virtue of the doxastic commitments of the person in question.
If I am right that C4 is correct, some important results follow about the methodological proposals discussed earlier. First, maximalism receives no support from this version of conservatism. Goldman asked rhetorically (in a passage quoted earlier) why, if we allow some beliefs, don=t we allow all? C4 provides an answer. We reject the maximalist strategy because not all beliefs are of equivalent epistemic values. Some beliefs may be thought to be shown true; others may not. C4 provides a way of distinguishing a subclass of our beliefs that are privileged in a certain respect, and thus provides an answer to the question why we can start with some beliefs and not others.
C4 is of no more help in defending Chisholmian particularism than it is in defending maximalism. On C4, there is no special reason to prefer particular judgments to general unless it just happens that only particular judgments are thought to be shown true. I can see no reason to think that this claim is in general true, hence it would seem that one who accepts C4 will not be a particularist.
C4 is kinder to the Rawlsian method of reflective equilibrium, though. Presumably, if S believes that p is shown true and also believes p, this belief will be one which S takes to be most obviously true, i.e., p will be one of the preferred beliefs on Rawls=s method. Thus, C4 fits better with Rawls=s method than either of the other two.
One more connection needs to be mentioned between this version of conservatism and other positions. In the introduction, I claimed that conservatism is not wedded to internalist views about justification. Yet, the defense I have given of conservatism is clearly internalistic. This, of course, does not show that internalism about justification is true; but what it might suggest is that no sort of purely externalist theory of justification can be adequate. In order to avoid this conclusion, one of two things must be shown. Either it must be shown that the structure of justification can be supported completely without appeal to conservative doctrines at all, or some alternative defense of conservatism (one without internalistic aspects) must be found. Since the major views regarding the methodology of theory construction seem to presuppose some version of conservatism, the first disjunct does not look promising. I do not believe the second disjunct is promising either, but I shall not argue that here. In any case, our results issue a challenge to pure externalists: either cleanse the structure of justification of conservatism, or find a non-internalist defense of the doctrine.
In spite of the initial implausiblity of conservatism, it is obviously important to the acceptability of conservatism, it is obviously important to the acceptability of many philosophical views that are not implausible. We have seen that there is a way to defend conservatism, and thus indirectly to defend some of these several philosophical views that depend on conservatism. In carefully evaluating what sort of conservatism doctrine is acceptable, we have seen that other philosphical views that require a conservatism doctrine are not supported by the conservatism doctrine we have found adequate. Thus, not only have we seen the defensibility of conservatism, our discussion has shown that certain views are in need of a different sort of support that can be provided by a proper version of epistemic conservatism.
NOTES
1. See, e.g., W. V. 0. Quine: 1960, Word and Object, MIT Press, Cambridge, Massachusetts, pp. 20C21; Gilbert 1-larman: 1973, Thought, Princeton University Press, Princeton, p. 159.
2. It may not be initially obvious that getting to the truth in the long run is not a purely epistemic justification for conservatism. The following consideration suffices to show that this is true. Suppose there is some proposition which, if we believe it, we will get to the truth later; and if we do not believe it, we will not. It is compatible with this description that we have overwhelming evidence that the proposition in question is false; and thus any reason for accepting it will have to be pragmatic or practical, and not purely epistemic. I explain the concept of purely epistemic rationality later in the paper.
3. W. V. 0. Quine: 1965, >Epistemology Naturalized=, in Ontological Relativity and Other Essays, Columbia University Press, New York, pp. 69-90.
4. Cf. Donald Davidson: 1975, >Thought and Talk=, in Samuel Guttenplan (ed.), Mind & Language, Clarendon Press, Oxford, pp. 20-21.
5. Recent discussion of the connection between epistemic rationality and truth can befound in Stewart Cohen: 1984, >Justification and Truth=, Philosophical Studies, 46,279-96; and in my >How to Be a Reliabilist=, American Philosophical Quarterly 23,1986.
6. The maximalist strategy is defended by Alvin Goldman: 1979, >Varieties of Cognitive Appraisal=, Nous 13; and by Keith Lebrer: 1974, Knowledge, Clarendon Press, Oxford. I borrow this discussion of the relationship between the maximalist strategy and conservatism from Richard Foley: 1983, >Epistemic Conservatism=, Philosophical Studies 43.165-82.
7. Cf. John Rawls: 1971, A Theory of Justice, Harvard University Press, Boston; Normal Daniels:1979, >Wide Reflective Equilibrium and Theory Acceptance in Ethics=, Journal of Philosophy 00, 256C82.
8. Chisholm=s particularism is articulated in Roderick Chisholm: 1977, Theory of Knowledge, 2nd ed., Prentice-Hall, Englewood Clifis, Chap. 7.
9. Thomas V. Morris: 1987, >Creatio ex nihilo: Some Considerations=, International Journal for Philosophy of Religion 14, 235.
10. Roderick Chisholm: 1981, Foundations of Knowing, University of Minnesota Press, Minneapolis, University of Minnesota Press, p. 14.
11. Chisholmp. 14.
12. 1 argue for this view in >Subjective Justification=, Mind, January 1984, pp. 71C84; Richard Foley defends the same view in >Contradictory and Near-Contradictory Beliefs=, in Peter French, Howard Wettstein, and Theodore Uehling, Jr. (eds.), Vol. X, University of Minnesota Press, Minneapolis, 1985.
13. Richard Foley, >Epistemic Conservatism=, p. 173.
14. Alvin Goldman: t986, Epistemology and Cognition, Harvard University Press, Cambridge, p. 1(17.
REFERENCES
Chisholm, Roderick: 1977, Theory of Knowledge, 2nd ed., Prentice-Hall, Englewood Clifis, New Jersey.
Chisholm, Roderick: 1981, Foundations of Knowing, University of Minnesota Press, Minneapolis.
Cohen, Stewart: 1984, >Justification and Truth=, Philosophical Studies, 46, 279C96.
Daniels, Normal: 1979, >Wide Reflective Equilibrium and Theory Acceptance in Ethics=, Journal of Philosophy, 256-82.
Davidson, Donald: 1975, >Thought and Talk=, in Samuel Guttenplan (ed.), Mind & Language, Clarendon Press, Oxford.
Foley, Richard: 1983, >Epistemic Conservatism=, Philosophical Studies, 43, 165C82.
Foley, Richard: 1985, >Contradictory and Near-Contradictory Beliefs=, in Peter French, Howard Wettstein, and Theodore Uehling, Jr. (eds.), Midwest Studies in Philosophy, Volume X, University of Minnesota Press, Minneapolis.
Goldman, Alvin: 1979, >Varieties of Cognitive Appraisal=, Nous, 13.
Goldman, Alvin: 1986, Epistemology and Cognition, Harvard University Press, Cambridge, Massachusetts.
Harman, Gilbert: 1973, Thought, Princeton University Press, Princeton, New Jersey.
Kvanvig, Jonathan: 1986, >How to Be a Reliabilist=, American Philosophical Quarterly,
Kvanvig, Jonathan: 1984, >Subjective Justification=, Mind, January, 71C84.
Lehrer, Keith: 1974, Knowledge, Clarendon Press, Oxford.
Morris, Thomas V.: 1983, >Creatio ex nihilo: Some Considerations=, International Journal for Philosophy of Religion, 14, 235.
Quine, W. V. 0.: 1960, Word and Object, MIT Press, Cambridge, Massachusetts.
Quine, W. V. 0.: 1969, >Epistemology Naturalized=, in Ontological Relativity and Other Essays, Columbia University Press, New York.
Rawls, John: 1971, A Theory of Justice, Harvard University Press, Cambridge, Massachusetts.
Department of Philosophy
Texas A & M University
College Station, TX 77843
U.S.A.