Underived Knowledge: An Answer to Skepticism about Induction and the External World Bryan Caplan Phil 122 Bojana Question #1 March, 1993 1. Introduction The problems of induction and the external world do indeed share a common structure. Both narrowly define what we "directly" or "immediately" know. They then note that we usually assume that we know much more than that, even though what we know directly does not logically imply most of what we assume we know. Therefore, much of our "knowledge" is unfounded. This paper will undertake three tasks. First, it will outline the general structure of this sort of skeptical argument. Second, it will show how the problems of induction and the external world fit exactly into this common structure. Third, it will point out the most questionable step in the two arguments -- which turns out to be the same in both cases. 2. The Common Structure of the Problems The problems of induction and the external world, as we shall see later, both fit elegantly into the following, more general skeptical argument: 1. X is known directly. 2. Y is not known directly. 3. X does not logically imply Y, even though we usually assume that it does. (4. Nothing else directly known implies Y either.) Therefore, we don't know Y. How does this argument work? Basically, it creates a "split" (in Prof. Stroud's words) between what we know directly and what we don't know directly. Hence, in order to know the latter, we would have to logically derive it from the former. Sometimes, this isn't a problem: even though I don't know that the Pythagorean theorem is true directly, I do know the properties of squares and triangles directly, and the Pythagorean theorem can be logically derived solely from knowledge of squares and triangles. But suppose that I don't directly know that something is true, and what I know directly does not logically imply it either. I would have to conclude that what I believed was knowledge was just an unwarranted assumption. 3. The Structure Applied I contend that this is exactly what happens with the problems of induction and the external world. In both cases, we start with a premise about what we know directly. Naturally, this is always tiny compared to the entirety of what we assume we know. Our next inclination is to try to deduce, to logically derive, the truth of our other beliefs from the small amount that we know directly. But upon inspection, we see that our other beliefs do not follow logically from what we know directly. So we don't have any reason to accept these beliefs. Consider how the problem of induction fits into the preceding structure. First, it is admitted that we have direct knowledge of what we currently observe. Second, it is claimed that we don't have direct knowledge of what we do not currently observe. Third, current observations never logically imply anything about the currently unobserved. As Hume explains, the argument would only work if we knew the law of cause-and-effect. But the latter is not known directly either, and to base it on induction would be circular. So even though we assume that we can base knowledge of the unobserved on our knowledge of the observed, we can't. Therefore, we have no reason to believe anything about unobserved matters of fact. (This argument is not actually complete. To make it valid, one would also need to know that we couldn't derive our knowledge of unobserved matters of fact from anything besides our knowledge of observed matters of fact. After all, even if X does not imply Y, another premise, X', could imply Y. Presumably, Hume did not consider this possibility because he took the distinction between relations of ideas and matters of fact to be exhaustive.) It will be easier to see how the problem of induction fits into the structure outlined in section two by stating the problem of induction more formally: 1. Currently observed matters of fact are known directly. 2. Currently unobserved matters of fact are not known directly. 3. Propositions about currently observed matters of fact do not logically imply propositions about currently unobserved matters of fact, even though we usually assume that they do. (We usually invoke the law of cause-and-effect, which is itself neither known directly nor logically deducible from aynthing known directly.) (4. Nothing else directly known implies any propositions about currently unobserved matters of fact.) Therefore, we don't know (i.e., have no reason to believe) anything about currently unobserved matters of fact. Now let us compare this to the problem of the external world. Following Descartes, most people will usually say that we directly know only that we have experiences of certain kinds. We directly know that "I feel as if I have a body" or "It looks as if I am seeing green." So what don't we directly know? We don't directly know anything about objects, about real, physical things. We don't directly know that "I have a body" or "I see green." To know "physical object" propositions at all, we would have to logically deduce them from knowledge of our experiences. But this is impossible. Since we could be dreaming or having delusions, the mere fact that I know that I am having a certain experience does not logically prove that real, physical objects exist. And again, assuming that nothing else known directly logically implies anything about the real, physical world, the conclusion that we never have any reason to believe anything about the physical world follows. Let us state the argument formally to see how it fits the general argument in section two: 1. Our experiences are known directly. 2. The real, physical world is not known directly. 3. Our experiences do not logically imply anything about the real, physical world, even though we usually assume that they do. (4. Nothing else directly known implies anything about the real, physical world either.) Therefore, we don't know (i.e., have no reason to believe) anything about the real, physical world. We are now ready to see that the structure of the problems of induction and the external world are exactly parallel at every step of the way. In both cases, premise #1 characterizes what we know directly in very narrow terms. In the problem of induction, it states that we directly know currently observed matters of fact; in the problem of the external world, it states that we directly know our experiences. In both cases, premise #2 states something crucial that we don't know directly. In the problem of induction, it states that we don't directly know anything about currently unobserved matters of fact; in the problem of the external world, it states that we don't directly know anything about real, physical objects. Then premise #3 blocks an inference from premise #1 to premise #2: in the case of induction, it says that we can't deduce anything about the currently unobserved from the currently observed; in the case of the external world, it says that we can't deduce anything about the external world from our experiences. Premise #4 just makes the argument valid by denying that there are alternative sources from which to derive premise #2. Finally, both arguments lead to parallel conclusions: in the first instance, we never have any reason to believe anything about currently unobserved matters of fact; in the second instance, we never have any reason to believe anything about the external world. Details aside, it turns out that the problems of induction and the external world share a common structure because they are both instances of the general argument spelled out in the second section. 4. A Common Solution The arguments sketched above -- the general one as well as its two applications -- are clearly valid. Given the premises, the skeptical conclusions follow inexorably. Hence, any solution would have to question the validity of the premises -- but which ones? In both cases, premise #1 is hard to doubt; and if we did doubt it, we would reach a deeper and more protracted skepticism, not the solution to the problems of induction and the external world. Premise #3 is likewise true in both cases. Propositions about the currently observed don't logically imply anything about the currently unobserved, and propositions about our experiences don't logically imply anything about the external world. Hume's and Descartes' thought experiments, in which the antecedent is true while the consequent is false, suffice to demonstrate this. Premise #4 is problematic: it is difficult to prove that no substitute premise exists, but unfortunately it is also difficult to positively produce a substitute premise. So premises #1,3, and 4 are unlikely escape routes from the problems of induction and the external world. What about premise #2? This one says that (in the case of induction) we don't have direct knowledge about currently unobserved matters of fact, and (in the case of the external world) that we don't directly observe real, physical objects. I don't think that either is true. While it would be hard to offer conclusive proof here, at least let me point out two plausible substitute premises in order to inspire doubt among convinced skeptics. Consider first this alternative premise about induction: The law of cause-and-effect can be directly known by reflective reason alone. We just turn our intellects to the law, and if we are sufficiently intelligent, honest, and open-minded, then we directly see its truth. The denial of the law isn't a contradiction, but perhaps reason does more than merely locate the presence or absence of contradictions.1 If this view is right, then premise #2 is false, since it states that we do directly know something about unobserved matters of fact -- namely, they will obey the law of cause-and-effect. And once we know, by reason alone, that the law of cause-and-effect is true, then we could argue from the law of cause-and-effect (major premise) plus knowledge of the observed (minor premise) to valid inductive conclusions. If this position is coherent, then we have one possible solution to the problem of induction. Now take the problem of the external world. Premise #2 states that we don't directly know anything about real, physical objects. Consider this plausible alternative: We directly see real, physical objects. (This view is called naive realism.) When I reflect on my observation, I realize that I don't see my experiences. I see real, physical objects. There is no inference involved. And if I do see real, physical objects directly, the "problem" of the external world collapses. John Searle, one modern naive realist, explains the theory succinctly: "Since I do not infer that there is a car there but rather simply see itÉit is not correct to say that the visual experience is the 'basis' in the sense of evidence or ground for knowing that there is a car there. The 'basis' rather is that I see the carÉ"2 Descartes would object that I might be dreaming; but if you could never know for sure that you directly saw the real world, how would you know that dreams existed in the first place?3 Once again, if this position is at least coherent then we have a viable solution to the problem of the external world. 5. Conclusion It is doubtful that anyone will be convinced in so short a space. Nevertheless, a solution like mine is the only likely escape route from skepticism. The other premises, as I noted, are very sturdy and hard to doubt. But isn't is plausible to think that we might have made a mistake in our initial characterization of what we directly know? When our premises about causality and perception imply absurd conclusions -- like total skepticism about induction and the external world -- we should re-examine them. When and if we do choose to re-examine them, we will find that a rationalist theory of the law of cause-and-effect plus a naive realist theory of perception are plausible alternatives that will never lead us to radical skepticism. Notes 1: As I argued in my previous paper, (Bryan Caplan, "An Enquiry Concerning Hume's Misunderstanding" unpub. ms.), it is quite possible that reason does more than locate the presence or absence of contradictions. And if it doe, then knowledge of the law of cause- and-effect based on reason alone is quite conceivable. 2: John Searle, Intentionality (Cambridge: Cambridge University Press, 1983), p.73. 3: If you find this one-line refutation implausible, let me elaborate. If you could never know for sure that you were right, then you couldn't know that you had ever made a mistake. And if you could never know that you were awake, you could never know that dreams existed. After all, only when you were conclusively awake and observing the real world could you know if there were really dreams. And similarly, if you could never know that an evil genius weren't deceiving you, then you could never know that, so all your claims about the evil genius would be unjustified too. In general, all of Descartes' skeptical arguments assume precisely what the argument denies, so they couldn't be valid.