You are here

10: Prospect and Retrospect

It is time to take stock. We noted in chapter I that twentieth-century British and American epistemology has been dominated by internalist notions, the most important of which is justification. We also noted that contemporary epistemology presents a vast, confused, and confusing welter of views, and that in two crucial respects. First, there is a great deal of confusion as to what the connection is between justification, on the one hand, and knowledge, evidence, and internalist constraints on the other. Second, there is the same confused and confusing welter of opinion as to what justification itself is. Among some of the more popular candidates: justification is taken as a matter of epistemic responsibility or aptness for epistemic duty fulfillment, as an “evaluation” of how well you have fulfilled your epistemic goals, as being believed or accepted on the basis of an adequate truth-conducive ground, as being produced by a reliable belief-producing mechanism, as being supported by or fitting the evidence, and as a matter of everything's going right, for the knowing subject, with respect to cognitive processes ‘downstream from experience’.

We saw that order can be introduced into this chaos by tracing the notion of justification back to its source in the classical deontologism of Descartes and (perhaps more important) Locke. Descartes and Locke speak of epistemic duty and obligation. Descartes seems to say that one is obliged not to believe anything that is not clear and distinct; Locke sees our main epistemic duty as that of proportioning belief to the evidence afforded by what is certain: my duty is to believe a proposition that is not certain for me to the degree to which it is probable with respect to what is certain for me. Now some of our contemporaries (BonJour, the classical Chisholm) explicitly explain justification in terms of responsibility or aptness for epistemic duty fulfillment, thus following the deontological lead of Descartes and Locke; others (Alston, Conee and Feldman, many others) explain justification as believing on the basis of evidence, that is, in terms of the content of what Locke sees as the principal epistemic duty. Still other views (Lehrer, Cohen) can be understood as related to that original deontological notion by way of analogical extensions of one sort or another. And in still other cases (Goldman), there is no real conceptual connection with that deontological notion, but a verbal connection instead: justification is used as a name for that quality or quantity, whatever it is, enough of which is sufficient to turn true belief into knowledge (and in Goldman's view that quality or quantity is not justification properly so-called). Finally, we saw that the internalist concerns characteristic of contemporary epistemology are also to be understood in terms of that original deontology; for internalism flows from deontology and is unmotivated without it.

We then turned to an examination of contemporary internalist accounts of warrant. What is characteristic of internalist accounts is that they see warrant as essentially a matter of justification: that quality or quantity that turns true belief into knowledge just is, according to internalists, justification, or perhaps justification together with a fillip to mollify Gettier. We began with the impressive work of Roderick Chisholm. There is a great deal to be learned from Chisholm; but what we saw was that justification, fulfillment of epistemic duty, is neither necessary nor anywhere nearly sufficient for warrant. As we also saw, both the classical and the post-classical Chisholmian accounts founder on the rock of epistemic malfunction. We noted that considerations of proper function also demonstrate the inadequacy of coherentism, whether taken neat or in the perceptive, sophisticated form put on offer by Laurence BonJour. Turning to Bayesianism, a specifically twentieth-century form of coherentism, we observed first that it has little to contribute to an account of warrant, being focused instead on that baffling, elusive, and pluriform notion of rationality; and a consideration of Bayesianism offered an opportunity to explore that notion. What the Bayesian offers us is a picture of one aspect of an ideally rational creature—or rather a picture of rationality (in the traditional sense of capability of thought, reflection, and inference) extended and idealized in one particular direction.

Failing to find a satisfying account of warrant among the unequivocal internalists, we turned next to the equivocal internalism of John Pollock, whose work occupies an interesting but uneasy halfway house between internalism and externalism. Fundamentally, according to Pollock, you are justified in believing a proposition A if you have arrived at it in conformity to your epistemic norms. The chief problem was that it seems to be perfectly possible to proceed in terms of incorrect norms, in such a way that one's beliefs would be Pollock-justified but still have no warrant. We then moved to the explicitly externalist and reliabilist accounts of Alston, Goldman and Dretske. Reliabilism is a substantial step in the right direction; perhaps we could see reliabilism as a zeroeth approximation to the truth of the matter. Still, it suffers from deeply debilitating problems—problems that need not be recounted here (since, after all, they appear only one chapter back) but which center, again, on the notion of proper function.

The idea of proper function figures prominently in the difficulties with the main current views of warrant; that suggests that this notion is much more deeply involved in our idea of warrant than is currently recognized. That suggestion, I think, is no more (or less) than the unvarnished truth. In Warrant and Proper Function I develop this idea in detail; here I shall briefly note its major features. What I propose to explain and explore is our notion of warrant, a notion nearly all of us have and employ in our everyday pursuits. This notion is not best explained, I think, just by producing a set of severally necessary and jointly sufficient conditions. Such a procedure is at home in logic and mathematics; it works somewhat less well or less directly in, say, the metaphysics of modality, and it works still less well (or still less directly) in epistemology. What we really have are paradigms: central, clear, and unequivocal cases of knowledge and warrant. But there is also a sort of penumbral zone of possible cases surrounding the central cases; these cases don't conform exactly to the conditions characterizing the central cases, but are instead related by way of analogical extension and similarity. And there is an even more shadowy belt of possible cases beyond that one, an area that is constituted by borderline cases, cases where it is not really clear whether what we have is knowledge (warrant) or not. More exactly, there are borderline cases between the central paradigmatic cases and those comprising the analogical zone, and borderline cases between the latter and cases that are not cases of warrant at all.

Accordingly, a good way to characterize our concept of warrant (more precisely: our system of analogically related concepts of warrant) is to specify the conditions governing the central paradigmatic core (here necessary and sufficient conditions are appropriate) together with some of the analogical extensions and an explanation of the analogical basis of the extension. This procedure is less elegant and regrettably more complex than the straightforward analysis we learned in our youth; it is certainly less stylish than setting out, in an austerely elegant clause or two, the necessary and sufficient conditions governing the concept. Following the procedure I advocate also makes the task of evaluating a proposed account much more difficult: is the putative counterexample a genuine counterexample to the conditions put forward as necessary and sufficient for the central and paradigmatic core concept? Or does it instead fall into one of those penumbral areas? All I can say by way of self-exculpation is first, we must heed Aristotle's dictum not to expect more clarity than the subject permits, and second, this procedure, if less elegant, is also more realistic and can take us closer to the truth.

The notion of proper function, I say, is crucial to the central paradigms of knowledge and warrant. But that notion is inextricably involved with another: that of the design plan of the organ or organism or system in question—the way the thing in question is supposed to work, the way it works when it works properly, when it is subject to no dysfunction. Design plan and proper function are interdefinable notions: a thing (organism, organ, system, artifact) is functioning properly when it functions in accord with its design plan, and the design plan of a thing is a specification of the way in which a thing functions when it is functioning properly. So we might say, as I did, that the central notion with respect to warrant is proper function; but we might as well say that the central notion is that of the design plan. In any event, the first condition for a belief's having warrant, as I see it, is that it be produced by faculties functioning properly. But this is by no means nearly sufficient. A second condition is that the cognitive environment in which the belief is produced must be the one or like the one for which it is designed. Your epistemic faculties may be functioning with perfect propriety; you may have passed your yearly cognitive examination at MIT with flying colors; but if you have been suddenly transported to a wholly different and alien cognitive environment, your beliefs may have little or no warrant. The basic picture, here, is that we have cognitive faculties that are adapted (by God or evolution or both) to our surroundings our cognitive environment; and when a belief is produced by these faculties functioning properly, then we have warrant.

But we do have to add another condition or two. First, not nearly every case of cognitive proper function is aimed at the truth, at the production of true beliefs. Suffering from a usually fatal disease, you may form unrealistic beliefs about the probability of your recovery; you may think that probability much greater than the statistics at your command would warrant. This need not be a case of epistemic malfunction; perhaps we are so designed that under such conditions we form beliefs more optimistic than the statistics warrant, because such optimism itself increases the chances of survival. (Compare William James's case of the climber whose confidence that he can leap the crevasse is too high from one point of view [he has never before leapt a crevasse of that size], but is also required if he is to have any chance at all of leaping it.) More generally, many beliefs are formed partly as a result of (for example) wishful thinking; and it is not at all clear that what we have there is epistemic malfunction. Wishful thinking has its purposes, even if forming beliefs with maximal verisimilitude is not among them. Beliefs of this sort, then, are produced by properly functioning cognitive faculties in the environment for which those faculties were designed; nevertheless they lack warrant. So a further condition must be added; to have warrant, a belief must also be such that the purpose of the module of the epistemic faculties producing the belief is to produce true beliefs. Another way to put it: the belief has warrant only if the segment of the design plan governing its production is aimed at truth, at the production of true beliefs.

Being produced by properly functioning faculties aimed at truth in the environment for which those faculties are designed—these are central aspects of our concept of warrant. There is one more central aspect of it, however: and that is that the design plan of the faculties in question must be a good one. An angel might design my faculties, aiming at producing a rational creature whose beliefs were for the most part true. If this angel is one of Hume's lazy or incompetent or immature angels, however, then the fact that my beliefs are produced by faculties functioning properly (that is, just as they were designed to) in the environment for which they were designed, and according to a design plan aimed at the truth—that fact will not be sufficient for warrant. It is also necessary that the design in question be a good design: that is, that there be a substantial objective probability that a belief of that sort produced under those conditions is true. We might call this the presupposition of reliability; it is the condition of warrant the reliabilist seizes upon. Although it is not the whole truth, it is indeed a part of the truth, and it is for this reason we may see reliabilism as an approximation (if only a zeroeth approximation) to the truth.

After outlining the central idea, I shall turn to a number of qualifications having to do with the design plan. There is functional multiplicity: the fact that a given bit of the design plan may be aimed at the fulfillment of more than one purpose; there is the distinction between the design plan and the max plan, between purpose and design, between intended results and unintended by-products, between responses that fulfill the main purpose of a system and those that are there as a result of trade-offs and compromises, and so on. (This last distinction gives us a way of handling an analogue of the dreaded Gettier problem.) Then in the next few chapters I shall explore some of the main areas of our epistemic establishment, turning successively to memory, knowledge of one's self, knowledge of other persons, perception, a priori knowledge and belief, induction, and probability. I next explore certain general features of the epistemic design plan: its foundationalist structure, the defeater and overrider system, and the place of evidence. In the final two chapters, I shall argue, first, that it is extremely difficult to see how to give anything like a naturalistic account or analysis of the notions of proper function, design plan, and their colleagues in that circle of interdefinable notions. I shall then conclude by arguing that while the view that I am proposing indisputably falls under the rubric of naturalistic epistemology, the latter flourishes much better in the garden of supernatural theism than in that of metaphysical or theological naturalism.