Hume’s Externalist Reply to the Problem of Induction



Introduction (as Abstract)

David Hume has long been regarded as a skeptic about induction. The basis for this widespread belief stems from Hume's attitude toward what has come to be called "the uniformity principle," the principle that states, roughly, that the future resembles the past. This principle is alleged to underlie all inductive inferences. Consequently, many claim that if this principle is unjustified, then every belief grounded on it is also unjustified.

D. C. Stove, Barry Stroud, and Galen Strawson are typical of those who regard Hume as a skeptic. Referring to the uniformity principle as the "Resemblance Thesis," Stove (1973) claims that Hume believed that "the Resemblance Thesis cannot be validly inferred from necessarily true premisses ... [and] any inductive argument for the Resemblance Thesis is circular" (45).[1] Consequently, Stove claims, Hume believed that "all predictive-inductive inferences are unreasonable" (34). Similarly, Stroud (1977) argues that Hume believed that "no one could ever reasonably believe anything about the unobserved on the basis of what has been observed" (54).[2] And, Strawson (1989) attributes the following view to Hume: our beliefs about causality "cannot be known to be true" (99).[3]

Increasingly, these days, philosophers are arguing that Hume is not a sceptic about induction. For example, Annette Baier (1991) argues that "if Hume really distrusts causal inference, and the inductions on which ... it rests, then he must distrust his own Treatise" (55). Similarly, David Owen (1999) argues that "it will not do to treat Hume as entirely negative about probable reasoning, if we are to avoid saddling him with a virtual contradiction at the heart of his enterprise" (146). And Don Garrett (1997) claims that Hume believes that "our evaluations of human reasoning skills ... presuppose the correctness of induction" (95).[4] But, the skeptical reading of Hume persists.

My aim in this paper is to critically analyze another version of the skeptical reading of Hume. In a recent article, Samir Okasha (2001) criticizes the argument Hume allegedly develops in support of skepticim about induction. Okasha's aim is to refute Hume's argument, and thus defend our knowledge gained by induction. He claims that Hume's argument rests on "a flawed description of our inductive behavior" (308). Consequently, Okasha believes that we should reject Hume's skeptical conclusion. My aim is to clarify Hume's intentions in his discussion of induction and show that it is a mistake to regard him as a skeptic about induction. Rather, I argue that Hume provides an externalist justification for induction. In addition, I want to show that Okasha's Bayesian solution to the problem of induction fails to take us beyond Hume's own solution to the problem of induction.

Hume’s Externalist Reply to the Problem of Induction

I. Introduction.

David Hume has long been regarded as a skeptic about induction. The basis for this widespread belief stems from Hume's attitude toward what has come to be called "the uniformity principle," the principle that states, roughly, that the future resembles the past. This principle is alleged to underlie all inductive inferences. Consequently, many claim that if this principle is unjustified, then every belief grounded on it is also unjustified.

D. C. Stove, Barry Stroud, and Galen Strawson are typical of those who regard Hume as a skeptic. Referring to the uniformity principle as the "Resemblance Thesis," Stove (1973) claims that Hume believed that "the Resemblance Thesis cannot be validly inferred from necessarily true premisses ... [and] any inductive argument for the Resemblance Thesis is circular" (45).[1] Consequently, Stove claims, Hume believed that "all predictive-inductive inferences are unreasonable" (34). Similarly, Stroud (1977) argues that Hume believed that "no one could ever reasonably believe anything about the unobserved on the basis of what has been observed" (54).[2] And, Strawson (1989) attributes the following view to Hume: our beliefs about causality "cannot be known to be true" (99).[3]

Increasingly, these days, philosophers are arguing that Hume is not a sceptic about induction. For example, Annette Baier (1991) argues that "if Hume really distrusts causal inference, and the inductions on which ... it rests, then he must distrust his own Treatise" (55). Similarly, David Owen (1999) argues that "it will not do to treat Hume as entirely negative about probable reasoning, if we are to avoid saddling him with a virtual contradiction at the heart of his enterprise" (146). And Don Garrett (1997) claims that Hume believes that "our evaluations of human reasoning skills ... presuppose the correctness of induction" (95).[4] But, the skeptical reading of Hume persists.

My aim in this paper is to critically analyze another version of the skeptical reading of Hume. In a recent article, Samir Okasha (2001) criticizes the argument Hume allegedly develops in support of skepticim about induction. Okasha's aim is to refute Hume's argument, and thus defend our knowledge gained by induction. He claims that Hume's argument rests on "a flawed description of our inductive behavior" (308). Consequently, Okasha believes that we should reject Hume's skeptical conclusion. My aim is to clarify Hume's intentions in his discussion of induction and show that it is a mistake to regard him as a skeptic about induction. Rather, I argue that Hume provides an externalist justification for induction. In addition, I want to show that Okasha's Bayesian solution to the problem of induction fails to take us beyond Hume's own solution to the problem of induction.

II. Okasha's Concerns.

In this section, I want to examine both the argument Okasha attributes to Hume, and Okasha's critical assessment of that argument.

Okasha takes Hume to be arguing as follows. Hume believes:

(P1) We follow a particular rule, the uniformity principle, when we reason about the future on the basis of past experience (Okasha 2001, 313). Specifically, we assume that the future will resemble the past.

But, Hume also believes:

(P2) The uniformity principle is unjustified. That is, there is no non-circular inductive justification for the uniformity principle, nor can the principle be justified deductively (311).

Consequently, Hume concludes:

(C) We are unjustified in our beliefs about the future that are based on past experience. More precisely, Okasha claims that Hume concludes that "past experience gives us no reason to believe anything about the future" (312).[5]

Okasha's criticism of Hume's argument for skepticism is two-pronged. First, he objects to Hume's assumption that we follow a rule when we make inductive inferences. As far as Okasha is concerned, this assumption is unsupported, and there is an alternative plausible account of inductive behavior that does not rely on it. Second, Okasha believes that because we do not rely on a rule when we make inductive inferences, Hume's argument is unsound. Consequently, Hume fails to establish his conclusion. Hence, we need not and should not be skeptical about induction.

Okasha grants that Hume's argument is persuasive. Still, Okasha insists that Hume's argument is unsound (308). The problem with Hume's argument is that it relies on a false premise, (P1). According to Okasha, an implicit assumption of (P1) is that our inductive behavior is guided by a rule. That is, in claiming that our inductive behavior is guided by the rule referred to as "the Uniformity Principle," Hume assumes that our inductive behavior is rule governed. Okasha argues that those who find Hume's argument for skepticism about induction persuasive uncritically accept Hume's assumption that there is a rule that guides our inductive behavior (314). But, Okasha believes that this assumption is false. Consequently, he objects to Hume's particular rule, the uniformity principle. If Okasha is correct about the fact that we do not follow a rule, then of course he is also correct in claiming that the rule that Hume allegedly identifies is not the rule we follow.

To prove that Hume is mistaken about our inductive behavior, Okasha presents an alternative account of our inductive behavior, one that he believes is more plausible than Hume's account. Okasha draws on the impressive resources of the Bayesians. He believes that, rather than relying on the uniformity principle, people update their beliefs on the basis of new experiences (316). As Okasha explains, "the basic way of learning from experience, according to Bayesians, is to conditionalize one's prior probability function on the deliverances of experience" (316).

Importantly, Okasha insists that Bayesian updating does not constitute "following a rule" (316). As he explains, "a rule of inductive inference is supposed to tell you what beliefs you should have, given your data, ... [but] the rule of conditionalization does not do that" (316). This is a subtle point, and crucial to Okasha's argument, so it is worth examining this point in detail. A rule, Okasha explains, determines an outcome. That is, a rule of inference determines what one believes. But, Bayesian updating does not work like this. When one updates in light of new experiences one may alter one's belief. But one is not necessarily led to alter one’s beliefs. Whether or not one alters one's belief in light of new data is a function of what one believed beforehand. That is, it is a function of one's prior probabilities. Consequently, two people confronted with the same new piece of evidence may react in different ways. For example, if Anna assigns a very low probability to the belief (R), "it will rain today," and a dark cloud appears, then she is apt to increase the probability assigned to (R). But, even after seeing the cloud she might still assign the belief a probability less than 0.5. Consequently, she might persist in believing that (R) is false. Alternatively, if Bernie assigns a higher prior probability to (R) than Anna does, though still a probability less than 0.5, the appearance of the dark cloud could lead him to assign a probability higher than 0.5 to (R). Consequently, after the appearance of the cloud Bernie would approach his day with the conviction that it is apt to rain.

?

More formally:

Bayes' Theorem Pr (q/p) = Pr (q) x Pr (p/q)

[Pr (q) x Pr (p/q)] + [Pr (~q) x Pr (p/~q)]

Pr (q/p) is the probability that it will rain given that there is a dark cloud in the sky. This is the probability Anna and Bernie wish to determine.

Pr (q) is the probability that it will rain, that is, the prior probability. Let us assume that Bernie's prior probability is 0.3 and Anna’s prior probability is 0.1.

Pr (~q) is the probability that it will not rain. For Bernie this is 0.7 and for Anna it is 0.9.

Pr (p/q) is the probability that there will be a dark cloud given that it rains. Let us assume that Bernie and Anna both give this value 0.8.

Finally, Pr (p/~q) is the probability that there will be a dark cloud given that it does not rain. Let us assume that Bernie and Anna both give this value 0.2.

Given these values, the sighting of the dark cloud will lead Bernie to assign a value of 0.63 to Pr (q/p) and Anna to assign a value of 0.31 to Pr (q/p).

?

The important point is that Bayesian updating can lead two people to hold different beliefs even when they are confronted with the same evidence. Hence, Bayesian updating does not constitute a rule of inference in the strict sense.

Given that our inductive behavior can be more adequately accounted for without assuming it is rule governed, Okasha believes that Hume's first premise, which assumes that our inductive behavior is rule governed, is false. Consequently, Hume's argument is unsound. Consequently, we need not be skeptical about induction.

III. Defending Hume.

In the remainder of this paper, I want to raise three concerns with Okasha's criticism of Hume. My aim is to defend Hume. First, I argue that Okasha has inaccurately characterized Hume's account of our inductive behavior, for Hume's Uniformity Principle is not aptly described as a rule of induction. Second, I argue that Okasha's Bayesian solution to the problem of induction is no better than Hume's own solution. Third, I argue that Hume is not a skeptic about induction. Let us now examine each of these concerns in detail.

My first concern is that, contrary to what Okasha suggests, Hume does not think that there is a rule of inference that guides our inductive behavior. In making my case, I cite two types of evidence. First, I examine textual evidence from Hume's Enquiry, the source of the argument that concerns Okasha. Second, I examine the nature of the Uniformity Principle.

Rather than attributing our inductive behavior to a rule, Hume attributes it to habit or custom. As Hume puts the point, "custom ... is the great guide of human life. It is that principle alone, which renders our experience useful to us, and makes us expect, for the future, a similar train of events with those which have appeared in the past" (29). Hume argues it is our nature to behave this way. As Hume explains, our habit of assuming that the future will resemble the past is "an operation of the soul ... as unavoidable as to feel the passion of love, when we receive benefit; or hatred when we meet with injuries. All these operations are a species of natural instincts, which no reasoning or process of thought and understanding is able, either to produce, or to prevent" (30).[6]

Hume argues that something as important to our survival as is inductive behavior would not be left to a faculty as unreliable as reason (37). That is, our species would not have survived if we depended upon reason. As Hume notes, reason is "slow in its operations; appears not, in any degree, during the first years of infancy; and at its best, in every age and period of human life, extremely liable to error and mistake" (37). Indeed, one of the things that Hume was desperately trying to aid his contemporaries in seeing is the limitations of reason. Hume claims that, like animals, we are chiefly moved by passions. Indeed, he notes that one of the reasons we cannot be persuaded by the arguments of the skeptics is because they fail to engage our passions (27). As Hume explains, "every passion is mortified by [skepticism], except the love of truth" (27). "By flattering no irregular passion, [skepticism] gains few partizans: By opposing so many vices and follies, it raises to itself abundance of enemies, who stigmatize it as libertine, profane, and irreligious" (27). Indeed, according to Hume, "nature will always maintain her rights, and prevail in the end over any abstract reasoning" (27). Thus, though as philosophers we may find skeptical reflections compelling, as humans such reflections have no power to alter our behavior.

There is additional evidence that Hume did not believe that our inductive behavior was rule governed. Okasha claims that the mark of a rule of induction is that it determines what one believes. But, contary to what Okasha implies, the Uniformity Principle would not determine what one believes. The habit that we have by nature of assuming the future will resemble the past does not determine what one believes. Guided by the uniformity principle, what one believes in the light of new evidence will depend upon what one has experienced in the past. Hence, two people guided by the uniformity principle encountering the same new experience could be led to hold different beliefs.

To illustrate this consider our earlier example, belief (R), "it will rain today." Recall that the sight of a dark cloud led Bernie to come to believe (R). Consider another person, Carol, who did not believe that (R) was true before the appearance of the dark cloud, who has never before encountered a dark cloud in her life, and who would not be led to believe (R) despite the fact that, like Bernie, she assumes the future resembles the past. Carol's past experience has not inculcated in her the habit of expecting rain when dark clouds appear. Thus, because the uniformity principle does not determine what beliefs one accepts, it is not aptly described as a rule of inductive inference in the strict sense.

?

More formally:

Bayes’ Theorem Pr (q/p) = Pr (q) x Pr (p/q)

[Pr (q) x Pr (p/q)] + [Pr (~q) x Pr (p/~q)]

Pr (q/p) is the probability that it will rain given that there is a dark cloud in the sky. This is the probability Carol wants to determine.

Pr (q) is the probability that it will rain, that is, the prior probability. Let us assume that Carol's prior probability is the same as Bernie's, 0.3.

Pr (~q) is the probability that it will not rain. For both Carol and Bernie this is 0.7.

Pr (p/q) is the probability that there will be a dark cloud given that it rains. Let us assume that Carol gives this the value of 0.5.

Finally, Pr (p/~q) is the probability that there will be a dark cloud given that it does not rain. Let us assume that Carol gives this the value of 0.5.

Given these values, the sighting of the dark cloud which led Bernie to assign a value of 0.63 to Pr (q/p) will lead Carol to assign a value of 0.3 to Pr (q/p). Hence, though Bernie's belief is changed in light of the dark cloud, Carol's belief is not changed. And it is not changed even though both she and Bernie accept the uniformity principle.

?

Let us consider the second concern I have with Okasha's critical analysis of Hume's analysis of induction. I believe that Okasha's Bayesian solution to the problem of induction gives us nothing more than Hume's own solution gives us. Hume claims that when we reason inductively, we uncritically assume the future will resemble the past. What Okasha fails to realize is that the Bayesians also make such an assumption. Consider the example of Anna and Bernie revising their beliefs in light of the appearance of the dark cloud. Each of them assumes that a dark cloud increases the chance of rain today. To take today's cloud as an indication of an afternoon rain storm is to assume that the future will resemble the past. That is, it is to assume that past dark clouds are indicators of the causal properties of today's dark clouds. Indeed, it seems that there would be no reason to assign a probability that deviates from 0.5 to any event occurring in the future if one did not take past experience as an indicator of what will happen in the future. There would be no grounds for assigning such a probability without the assumption that the future resembles the past.

Indeed, this is Hume's point. All inductive inferences are ultimately grounded on the assumption that the future resembles the past. Now this does not mean that Okasha's Bayesian solution to the problem of induction fails. It merely suggests that Bayesians are unable to ground inductive inference on anything more secure than that on which Hume grounded it.

Let us now consider my third concern. Contrary to what Okasha claims, Hume is not a skeptic about induction. Again, in making my case I will cite two types of evidence. First, close scrutiny of Hume's text does not support the skeptical reading. Second, Hume actually suggests that inductive inference are justified, albeit not in the typical way that people think of induction.

I believe that Okasha fails to see that Hume is not a skeptic because he has based his reconstruction of Hume's argument on a misreading of Section IV of the Enquiry. What Hume claims in Section IV is that beliefs about things not present to the senses or memory cannot be grounded a priori. In his words, "it is not reasoning which engages us to suppose the past resembling the future, and to expect similar effects from causes, which are, to appearance, similar" (25).

Hume is explicit that his goal in Section IV is quite modest. He merely aims to give a negative answer to the question: "What is the foundation of all conclusions from experience?" (20). That is, he merely aims to tell us what is not the foundation for our conclusions from experience. And, his answer is that "our conclusions from experience are not founded on reasoning, or any process of the understanding" (21). Okasha, though, mistakenly attributes to Hume the view that we are unjustified in our inductive behavior. To support the skeptical conclusion that Okasha attributes to Hume, one would have to claim much more than Hume claims. But, nowhere does Hume claim that past experience gives us no reason to believe anything about the future, as Okasha suggests. Nor does Hume accept any of the other three skeptical claims attributed to him earlier in the Introduction of this paper. That is, Hume does not claim that all predictive-inductive inferences are unreasonable; nor does he claims that no one could ever reasonably believe anything about the unobserved on the basis of what has been observed; nor does he claim that our beliefs about causality cannot be known to be true. In fact, Hume suggests that it would be quite irrational not to hold such beliefs (27). For these reasons, I believe that it is a mistake to regard Hume as a skeptic about induction.

Further, there is good reason to think that Hume believes that inductive inferences are epistemically grounded. One could fail to see this if one were an internalist about justification, that is, if one believed that the only justification for a belief is another belief or set of beliefs (see, for example, BonJour 1978). Hume, though, is an externalist about justification with respect to inductive inferences. An externalist believes that one can be justified or warranted in holding a particular belief even if one cannot cite another belief to justify it. One can be justified even when one lacks a supporting belief provided the belief is acquired by a reliable belief-forming process (see Goldman 1979). That Hume is an externalist is evident from the following passage in which he compares our inductive behavior to our kinaesthetic behavior.

As nature has taught us the use of our limbs, without giving us the knowledge of the muscles and nerves, by which they are actuated; so has she implanted in us an instinct, which carries forward the thought in a correspondent course to that which she has established among external objects; though we are ignorant of those powers and forces, on which the regular course and succession of objects totally depends. (Hume 1977, 37)

Thus, rather than being grounded by another belief, one's inductive inferences are ultimately grounded on the habit of assuming the future will resemble the past. Even though one cannot justify this assumption by the other beliefs one holds, Hume is confident that it is a reliable assumption. Indeed, according to Hume, inductive inferences are grounded on a foundation more secure than reason (37). Just as it would be foolish to think that we do not know how to move our arms because we do not understand how the muscles are moved when we move our arms, it would be foolish to think that inductive inferences are unjustified because we can cite neither an inductive nor a deductive proof for the assumption that underlies such inferences.

IV. Concluding Remarks.

In summary, I have argued that Okasha has misrepresented Hume's argument and view. Hume does not believe that we are guided by a rule when we reason inductively. Indeed, Hume has no intention to establish the conclusion that Okasha attributes to him, for Hume is not a skeptic about induction. Further, Hume does not think that beliefs arrived at inductively are ungrounded. What he is skeptical about is the power of reason to ground such beliefs. But, as an externalist, it is not necessary to ground our beliefs on other beliefs.

Further, I have argued that Okasha has not delivered what he has promised. He has not provided us with an alternative superior account of inductive behavior. His Bayesian account presupposes that we assume the future resembles the past. In this respect, Okasha's solution to the problem of induction is no better than Hume's own solution.

REFERENCES

Arnold, N. S. 1983. "Hume's Skepticism about Inductive Inferences," Journal of the History of Philosophy, 21: 1, January, 31-55.

Baier, A. 1991. A Progress of Sentiments. Cambridge: Harvard University Press.

Beauchamp, T., and A. Rosenberg. 1981. Hume and the Problem of Causation. Oxford: Oxford University Press.

BonJour, L. 1978. "Can Empirical Knowledge Have a Foundation?," American Philosophical Quarterly, 15: 1, 1-13.

Broughton, J. 1983. "Hume's Skepticism about Causal Inferences," Pacific Philosophical Quarterly, 64: 1, January, 3-18.

Garrett, D. 1997. Cognition and Commitment in Hume's Philosophy. Oxford: Oxford University Press.

Goldman, A. 1977. "What is Justified Belief," in G. S. Pappas, (ed.), Justification and Knowledge. Dordrecht: Kluwer Publishers, 1-23.

Hume, D. [1748]/1977. An Enquiry Concerning Human Understanding, edited by E. Steinberg. Indianapolis: Hackett Publishing Company.

Lipton, P. 1991. Inference to the Best Explanation. New York: Routledge.

Okasha, S. 2001. "What Did Hume Really Show about Induction?," Philosophical Quarterly, 51: 204, 307-327.

Owen, D. 1999. Hume's Reason. Oxford: Oxford University Press.

Reichenbach, H. 1938. Experience and Prediction. Chicago: University of Chicago Press.

Salmon, W. DATE. "The Justification of Inductive Rules of Inference," in I. Lakatos, (ed.), The Problem of Inductive Logic.

Stove, D. C. 1973. Probability and Hume's Inductive Scepticism. Oxford: Clarendon Press.

Stroud, B. 1977. Hume. London: Routledge and Kegan Paul.

Wright, J. P. 1983. The Sceptical Realism of David Hume. Minneapolis: University of Minnesota Press.

ENDNOTES

[1] For Stove (1973) the Resemblance Thesis is the claims that "unobserved instances will resemble observed ones" (34).

[2] Stroud's (1977) version of the uniformity principle comes directly from Hume, and it asserts that "instances, of which we have had no experience, must resemble those, of which we have had experience, and that the course of nature continues always uniformly the same. (p.89)" (54)

[3] Indeed, many others read Hume as a skepic about induction, including Lipton 1991, Reichenbach 1938, Salmon DATE, and Wright 1983.

[4] For additional examples of philosophers who say Hume is not a skeptic about induction see Beauchamp and Rosenberg 1981, Arnold 1983, and Broughton 1983.

[5] I have condensed Okasha's argument. His presentation involves seven claims, rather than the three I identify. His first claim corresponds to (P1). His second to fifth claim correspond to (P2). And his sixth and seventh claim correspond to (C), above.

[6] Garrett (1997) concurs with my reading. As Garrett explains, Hume "is arguing that we do not adopt induction on the basis of recognizing an argument for its reliability, for the utterly sufficient reason that there is no argument ... that could have this effect" (92).

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download