How to be a Pragmatist
C. I. Lewis and Humean Skepticism
Murray G. Murphey's masterful treatment of C. I. Lewis's philosophy leaves two things amply clear: first, that Lewis struggled with skeptical arguments from Hume throughout his career; and second, that Lewis never adequately resolved the problems raised by those arguments. In this paper I will consider Lewis's approach to Hume's skepticism in Mind and the World Order (MWO) and in An Analysis of Knowledge and Valuation (AKV), and I will argue that Lewis's reply to Hume in these works did not change as dramatically as Murphey claims.1 Nevertheless, I agree with Murphey that there are two quite different lines of reply discernable in Lewis, and that neither adequately answers Hume. In the final part of the paper I argue that Lewis's pragmatism gives us resources for an adequate reply to Hume's skeptical arguments, although it is not the reply that Lewis himself gives.
1. A Skeptical Problem from Hume
The broad outline of Hume's skeptical reasoning is well-known.2 On the one hand, knowledge seems to require assumptions about the reliability of our evidence: we must assume, for example, that our evidence makes our conclusions likely. On the other hand, we seem to be in no position to justify such assumptions. The problem can be raised locally or globally. Locally, we can ask whether some particular body of evidence E makes likely some particular judgment that P is the case. That is, in order for E to justify P, it seems that we require some assumption P', to the effect that E makes P likely. But it is not clear how the justification of P' is possible. For presumably P' is itself an empirical proposition, requiring empirical evidence E', and some assumption P" to the effect that E' makes P' likely. Call propositions of the form e makes p likely "connecting propositions." Considered locally, Hume's problem is the [End Page 24] problem of showing how any particular connecting proposition can be justified without falling into either vicious regress or vicious circles.
One might think that, in general, connecting propositions can be justified by the appropriate use of induction. But this is to raise the same problem globally. In general, it seems that inductive evidence warrants its conclusions only if we can assume R, that there is a regularity to nature. If the world exhibits no such regularity or uniformity, Hume points out, then past experience provides no evidence for conclusions about future experience.
Call R "The Regularity Principle," since it says that there must be some regularity or uniformity in nature. It is not clear how we are to justify R without falling into vicious regress or vicious circularity.
Since the global problem is the more fundamental, it will be helpful to state its structure more clearly.
Hume's Skeptical Argument
- Our judgments about unobserved matters of fact depend for their evidence on both a) observed cases, and b) some assumption R to the effect that there is regularity in nature. A different way to state R: Observed cases are a reliable indication of unobserved cases.
- But there is no way to justify R.
- Our judgments about unobserved matters of fact depend for their evidence on an unjustified assumption. (1, 2)
- Knowledge cannot depend for its evidence on an unjustified assumption.
- We have no knowledge of unobserved matters of fact. (3, 4)
Clearly, premise 2 of the argument is crucial. Hume's reasoning for premise 2 is formidable, however. Hume considers various sources of evidence for R's justification, and concludes that none can do the job. For example, R is itself a contingent proposition about the world--it says that the world is in fact uniform and that our experience will be in fact regular. But then various kinds of a priori evidence are irrelevant to this sort of proposition. In Hume's terminology, the proposition expresses a matter of fact rather than a relation of ideas. In Lewis's terminology, the proposition is synthetic rather than analytic. As such, the proposition must be justified by empirical evidence rather than by a priori evidence. But any empirical evidence, Hume argues, would be circular, requiring the very assumption that is in question. And so there are no sources of evidence for R's justification. Again, it is helpful to see the structure of Hume's argument more clearly.
Hume's Supporting Argument
- If assumption R is to be justified then it is by either a priori evidence or empirical evidence. [End Page 25]
- R cannot be justified by a priori evidence because R is contingent, and a priori evidence is relevant only to propositions that are necessary.
- R cannot be justified by empirical evidence because all empirical evidence must assume R, and so any empirical evidence would be viciously circular as evidence for R.
- R cannot be justified.
Hume's skepticism about induction is formidable--over two hundred years later, there is no consensus among philosophers about where Hume's reasoning goes wrong. But things are even worse for Lewis than for most, on two counts. First, on Lewis's account something like inductive reasoning is implicated in all of our knowledge about the world, and not just in our knowledge of unobserved matters of fact. For Lewis, then, even our perceptual knowledge of material objects is called into question by Hume's argument.
Second, Lewis's account of empirical knowledge raises a similar problem about memory. All of our empirical reasoning, Lewis observes, depends for its evidence on present but also past observations. And so all of our empirical reasoning requires a second assumption about the reliability of memory. (AKV 327) Call this assumption "The Memory Principle." Skeptical arguments analogous to those directed at R apply to The Memory Principle as well--there seems no way to justify that principle without falling into vicious regress or vicious circularity.
2. Lewis's Reply in Mind and the World Order
Lewis accepts something like Hume's distinction between matters of fact and relations of ideas. In Lewis's framework, there are contingent matters of fact justified by empirical evidence and there are analytic truths justified by a priori evidence. What is more, for Lewis these categories are exhaustive--there are no synthetic a priori truths. It is odd, then, that Lewis would attempt an a priori defense of the Regularity Principle. But this is exactly what he does in MWO. In effect, Lewis sets out to deny premise 2 of Hume'sSkeptical Argument and premise 2 of Hume'sSupporting Argument.
Lewis announces his intention as early as the Preface (MWO xxi), and again in Chapter Two: "any conceivable experience will be such that it can be subsumed under concepts, and also that predictive judgments which are genuinely probable will hold of it." (MWO 38) Lewis's argument for this thesis trades on two ideas: that conceiving essentially requires concepts, and that concepts essentially imply order and regularity. Accordingly, it is impossible to conceive of any experience that entirely lacks order. Here is Lewis:
What is required in the way of order if experience is to be intelligible and knowledge possible is only that there should be apprehensible things and objective facts--and to this we can conceive no alternative whatever, unless it be the non-existence of everything.
This assumption of the existence of things, that is, of certain recurrent correlations in the sequence of possible experience, is all that is required for the validity, as probable, of empirical generalizations or "laws," and of the argument from past to future with respect to these.
But why can't we give up this "assumption of the existence of things," the assumption that there are "apprehensible things and objective facts"? Lewis argues that this is impossible.
That conception in general should be invalid, is quite impossible. That attempt to envisage an experience or state of affairs such that every attempt to discover stabilities must fail, is the attempt to conceive the inconceivable--to conceive what would not be things or objective facts nor subject to any generalization which makes what is denoted conformable to concepts. The experience or reality which should be incompatible with conception, ipso facto cannot be conceived.
It seems that we must read Lewis's argument as follows:
- Necessarily, conceiving anything at all requires the application of concepts to experience. (From the definition of "conceiving.")
- Necessarily, the application of any concept to experience requires our conceiving some order or other, some stability or other, in the experience to which the concept is applied. (From Lewis's theory of concepts.)
- Necessarily, conceiving anything at all requires conceiving some order or other, some stability or other. It is impossible to conceive of any experience that entirely lacks order.
But read in this way the argument smacks of sophistry. The conclusion is now about the meaning of "conceiving" rather than about the world. Certainly, Lewis has not established the Regularity Principle or anything like the Regularity Principle.
It is fair to say that Lewis's strategy against Hume in MWO was doomed to failure from the start. Put simply, the strategy requires that Lewis treat the Regularity Principle as a necessary truth, and therefore as an analytic truth. But as Hume well saw, the principle is neither necessary nor analytic. That the world is regular rather than chaotic is a contingent fact about the world, and not a necessary truth about words.
3. Lewis's Reply in An Analysis of Knowledge and Valuation
Murphey tells us that Lewis has changed his views about probability in AKV, and that with these changes come a different reply to Hume. Specifically, Murphey argues, Lewis has now adopted Reichenbach's pragmatic approach to the problem of induction. Reichenbach's central idea is that our [End Page 27] faith in induction has a kind of practical rationality: we are justified in employing good inductive reasoning because, if that does not work, then nothing will. Reichenbach's idea is attractive: If there are various courses of action possible, none of which guarantees success but one of which holds out the possibility of success, then it is rational to embark on the latter. Murphey tells us that in AKV Lewis takes over this approach. Lewis recognized "that no such proof of induction as he had attempted in Mind and the World Order was possible. Lewis's answer here was fully pragmatic--if there is a right answer to be found, Reichenbach's method will find it, and if there is not no other approach will do any good" (290).
It is not clear to me, however, that Lewis's approach in AKV is so single-minded. Consider the following line of argument, from a section that Lewis calls "'Deduction' of the basic validity of memory and of induction."
Our sense of cumulative temporal experience, mnemonically presented within the epistemological present... is something of which we cannot divest ourselves; it is constitutive of our sense of the only reality by reference to which empirical judgments could have either truth or falsity or any meaning at all....
Empirical reality does not need to be assumed nor to be proved ... Nor does the thesis that empirical reality can be known, require to be postulated or to be demonstrated: it is an analytic statement which can only be repudiated on premises which already imply it. It is by overlooking this fact that the skeptic must always fall into contradiction.
What we would here maintain is that without genuinely knowable past experience, or without genuine relevance of past experience to the future, we could have no such sense of empirical reality.
Lewis is still trying to pull the rabbit out of the hat. The trick did not work in MWO, and it will not work here either, for the same reasons.
Nevertheless, Murphey is right that Lewis has an additional line of reply to Hume in AKV. We are justified in employing induction in that, if this does not work, nothing will. The problem with this sort of reply is clear, however. At most, it establishes that our inductive reasoning and our inductive beliefs have a kind of practical rationality. But the skeptic does not deny that our beliefs have practical rationality. Certainly Hume does not deny this. Rather, his point is that our beliefs are not well grounded in an epistemic sense. That is the point of Hume's Skeptical Argument above, and nothing in this "pragmatic" response from Lewis touches the reasoning of that argument.
4. A Better Reply to Hume's Argument.
We may now revisit Hume's Skeptical Argument to see that a better reply is available. Specifically, premise 1 of that argument is as follows:
- Our judgments about unobserved matters of fact depend for their evidence on both a) observed cases, and b) some assumption R to the [End Page 28] effect that there is regularity in nature, observed cases are a reliable indication of unobserved cases.
Premise 1 implies that the Regularity Principle (R) functions as part of ourevidence for judgments about unobserved matters of fact. It is this implication that generates the requirement that the Regularity Principle, or some such principle, be justified if our beliefs about unobserved matters of fact are to be justified. Much of contemporary epistemology would judge that this is a mistake, however. Justification and knowledge require not that the Regularity Principle is justified, but that it is true. So long as the principle is true, that is, inductive reasoning will be reliable and hence capable of generating justified belief and knowledge.
This is the answer that "reliablist" theories give to Hume's problem.3 The reliabilist answer is often met with disapproval, however. Typically, objections take two forms. The first is a kind of "internalism" with respect to epistemic evaluation. De facto reliability, it is argued, is insufficient for justification and knowledge. What is required is that one knows that one's evidence is reliable, and that one knows it from an "internal" perspective. One must be able to establish reliability "from the inside," so to speak, where this means doing so on grounds that are accessible by reflection alone, rather than on grounds that depend on further empirical investigation. The second objection to reliabilism is closely related. Specifically, it is argued that one's understanding of one's reliability cannot be "question-begging." It would not be adequate, for example, to simply assume that one's evidence is reliable. Rather, one must be able to establish this, and in a way that does not already take for granted that inductive evidence or inductive reasoning is reliable.
It is here, I believe, that Lewis's pragmatism gives us resources for answering the skeptic. Specifically, Lewis's pragmatism gives us resources for explaining why these latest skeptical demands are inappropriate.
Consider the first demand--that one knows that P only if one knows from an internal perspective that one's evidence for P is reliable. This is an exceedingly difficult demand to meet, and it plays a central role in the skeptical problematic reviewed above. But why should we accept it? Why should knowledge require that this sort of demand can be met? Lewis reminds us that knowledge is for action: "But knowledge has a practical business to perform, the interests of action which it seeks to serve." (MWO 238) Again, "Knowledge, action, and evaluation are essentially connected. The primary and pervasive significance of knowledge lies in its guidance of action: knowing is for the sake of doing." (AKV 3)
More specifically, knowledge guides action by allowing prediction and control.
Action attempts to control future experience, so far as may be, in our own interest. It has a terminus a quo in the situation which is given; its [End Page 29] terminus ad quem in some experience to which a positive value ... is assigned. The principal function of empirical knowledge is that of an instrument enabling transition from the one to the other; from the actual present to a future which is desired and which the present is believed to signalize as possible. To know is to apprehend the future as qualified by values which action may realize; and empirical knowledge is essentially utilitarian and pragmatic.
In many ways Lewis is sympathetic with the internalist approach to justification and knowledge. This is evidenced both by his demand that knowledge be grounded in "the given" of experience, and by his insistence that evidential relations can be known a priori. But Lewis's pragmatism clearly cuts against these internalist demands. If knowledge is for action, then knowledge should require no more than action requires. And it is hard to see why action requires any such thing as the internalist wants. In particular, it is not clear why one needs to know from the inside that one's evidence (or reasoning, or methods) are reliable. Why isn't it enough that they are reliable in taking us from the "terminus a quo in the situation which is given" to the "terminus ad quem in some experience"? Of course, it is preferable and sometimes necessary to verify the reliability of our cognitive activities. But the internalist places narrow restrictions on what sort of verification is allowable--it must be non-circular (or non-question-begging) in the very specific sense described above. But again, why should the purposes of action require that?
On behalf of the skeptic, one might answer as follows: Knowledge plays not only a practical role, but a social role as well. Knowledge is not merely for action, but also for the coordination of action. And this function demands that we can give each other our reasons, that we can verify our beliefs and the reliability of our methods to each other. This, one might argue, is the real source of the internalist demands above.
Lewis agrees that knowledge coordinates action, and that this practical function gives knowledge a social dimension.
Whoever knows or claims to know must admit of the pertinence of the challenge, "How do you know; what warrants you in believing?" And he must also find answer to the even more fundamental challenge, "What do you mean; and how will what you indicate disclose itself?" Implicitly he agrees that he should recede from his assertive attitude if either of these two challenges cannot be met.
But for Lewis, these requirements are by nature practical, and so are no stronger than our purposes require. Lewis continues as follows:
Even in the best and clearest cases of knowledge, such as are likely to be put forward as examples, our sense of what is meant, and our sense of the basis of belief, will be incomplete.... The utmost we can demand is that one who is said to truly know should be able to provide such explication [End Page 30] when the need of it is genuine, and after reflection, and up to a certain point--the point where we reach what is already understood or what may be taken for granted.
In these passages Lewis is acutely aware that knowledge plays a normative role in our thought and action. "Knowledge is not a descriptive but a normative category: it claims correctness; mental states are classified as genuine knowing only on the assumption of such correctness." (AKV 10) But the sort of normativity involved has its source in knowledge's practical function. As such, it is wrongheaded to demand more than that practical function requires.
To put things starkly, the dispute between the skeptic and the non-skeptic comes down to this: Both agree that knowledge is normative, but disagree about the requirements to which epistemic normativity gives rise. The skeptic claims that such-and-such requirements cannot be met; the non-skeptic insists that such-and-such requirements are inappropriate. How to resolve the dispute? My suggestion is that we take seriously Lewis's insight that justification and knowledge have social and practical functions. If these functions do not ground a particular requirement on justification and knowledge, then it is reasonable to conclude that no such requirement exists.
1. Mind and the World Order: Outline of a Theory of Knowledge (New York: Dover Publications, 1929), hereafter MWO; An Analysis of Knowledge and Valuation (LaSalle, Ill: Open Court Publishing, 1946), hereafter AKV.
2. Hume's arguments are found in A Treatise of Human Nature, second edition, L. A. Selby-Bigge, ed. (Oxford: Clarendon Press, 1978); and Enquiries Concerning Human Understanding and Concerning the Principles of Morals, third edition, L. A. Selby-Bigge, ed. (Oxford: Clarendon Press, 1975).
3. See, for example, James Van Cleve, "Reliability, Justification, and the Problem of Induction," Midwest Studies in Philosophy 9 (1984): 55567.