Lecture Notes, UC Davis Philosophy 102, Theory of Knowledge: Internalism vs. Externalism

Lecture Notes, UC Davis Philosophy 102, Theory of Knowledge

Internalism vs. Externalism

G. J. Mattey, Senior Lecturer

In the last few weeks, we have been looking at various accounts of knowledge, including foundationalism, coherentism, causal accounts, and reliabilism. Roughly speaking the first two sorts of account have been called "internalist" while the latter two have been called "externalist.' We have not yet seen any clear definition of what makes an account either "internalist" or "externalist."

Today's reading, from a paper by University of Iowa professor Richard Fumerton, is a look at various ways in which the distinction might be understood. Fumerton himself is an internalist, and he wants to be able to understand the distinction in a way that illuminates the basic motivation for adoption of each of the two sides.

Although the debate between internalists and externalists is central to contemporary epistemology, it is conducted under unsatisfactory circumstances. It is unclear what the two positions actually are. "It seems to me that philosophers are choosing sides without a thorough understanding of what the respective views entail."

There are several different ways of technically defining the difference. It is likely that there is no single "correct" way, as with most philosophical distinctions. One way to get at least a good distinction is to make sure that most philosophers who have been labeled internalists fall under the "internalist" side, and most who have been labeled externalists fall under the "externalist" side. But this is not Fumerton's main goal.

Instead, Fumerton is concerned to discover a distinction that is fundamentally important and that reveals why internalists have rejected externalist accounts of knowledge and justification. Or at least he wants to show why he rejects them, and at the same time to showcase his own brand of internalism, which is an "extreme version of foundationalism" which "has been neglected too long."

Perhaps the most obvious way of understanding internalism is as the view that justification or knowledge is an internal state of a believer or knower. Then externalism would be the position that justification or knowledge involves more than an internal state: it involves an external factor as well. Thus a causal account of knowledge might serve as a paradigm of externalism, in that in order to know that p, one must be caused to believe that p by the fact that p. But the fact that p is frequently external to the believer or knower. The fact that it is dark outside as I write this is surely external to my present cognitive state.

A test of internalism then would be whether two people in exactly the same internal state could differ with respect to whether they are justified or know. If in broad daylight someone has a dream just like my waking state, for example, it might be said by an externalist that such a person lacks justification and knowledge, though I have it in my present state.

This test would work with the causal account of knowledge, but it does not work for many well-known "externalist" theories. Let us take Goldman's reliabilism as an example. For him, what is crucial is the "cognitive process" (or, by extension, cognitive method) that produces a belief (our text, p. 398). A cognitive process or method is, on his view, a purely internal process.

The property "being caused by a process which satisfies a certain description" (here, the description is that it is a reliable belief-forming process) is a relational property, but it is still an internal property of the subject. Other externalists do not restrict the processes to internal ones. But they could still describe internal states in relational terms, such as the state of "being caused by a process which satisfies a certain description." This is a property of an internal state which makes external reference.

A way to preserve the distinction would be to say that for an internalist, the only states relevant to knowledge and justification are non-relational states. But Fumerton points out that an internalist taking this route would be very lonely. As far as knowledge is concerned, internalists adopt a truth condition. Two people in the same non-relational internal state can differ with respect to whether their beliefs are true. So almost everybody would be an externalist with respect to knowledge, on this account.

You might think that the situation is different with respect to justification. I and the dreamer might be thought to be equally justified in our beliefs about its being night, given our non-relational internal states. In that case, the test for internalism is passed. But there are accounts of justification that have been called "internalist" that would not pass the test.

Fumerton singles out as an example a kind of internalist view that will loom large in what follows. Philosophers such as Bertrand Russell (and himself) have held that we have non-inferential justification when we are directly acquainted with facts. This involves a relation of a person to a fact, which may be external to the person. It might be argued that for this reason, Russell's view is externalist. But Fumerton notes that "at least some externalists take such positions to be paradigms of the sort of internalist epistemology they are rejecting."

Another case is that of Descartes, "the externalist's favorite internalist." You will recall that for Descartes, the relation of "having an idea" is fundamental to human cognition. One has a justified belief when one has a clear and distinct perception of an idea. Even though an idea is internal to the person, the state of having an idea is, from a metaphysical standpoint, a relational state.

It seems that the only kind of purely non-relational internalism left is an "adverbial theory." According to this theory, perceptual states do not involve a relation to a sense-datum, idea, or physical object. Rather, they are just a way of perceiving. For example, one might be justified in believing that one is "seeing red-ly" in a situation where one actually is seeing red-ly. There need be no relation to anything red for one to have this kind of knowledge. But Fumerton rightly notes that this is a very narrow view not held by most internalists.

There seems to be a simple way to broaden the field for internalism. As we have seen in the case of Descartes, if we identify justification with believing on the basis of clear and distinct perception, then there is a relation involved, but the relation is to something which is itself internal, i.e., ideas. So the internalism would identify knowledge and justification "either with nonrelational properties of the mind or with relational properties of the mind where the relata of the relations are the mind and its nonrelational properties." This would at least broaden the scope of internalist accounts of justification, if not for knowledge (for reasons discussed above).

Fumerton has two objections to this move. First, the account of internalism is still narrow enough to exclude those who, like himself, think that we have non-inferential acquaintance with extra-mental facts. These facts may be "sense data" (construed in a non-psychological way), surfaces of physical objects, Platonic forms, or universals. Yet he is reluctant to call those who hold such views externalists because "I don't think paradigm externalists want their company."

The second objection is related to the first. There is a structural similarity between an account like Descartes's and one like Russell's and his own. Both think that there is non-inferential justification based on direct acquaintance. The only difference is what stands as the object of acquaintance: something mental or something non-mental. The two views are "fundamentally alike." The internalism/externalism controversy "will not be getting at a significant issue if one of these views gets described as a form of internalism while the other is described as a form of externalism."

A second attempt at defining the distinction is through the way in which we satisfy the epistemic conditions for justification or knowledge. The internalist could be seen as holding that in order to be justified in believing that p, we must be justified in believing that the conditions for justified belief that p have been satisfied. Similarly, in order to know that p, one must know that the conditions for knowing that p have been satisfied. Justification and knowledge, on this account, require for the internalist higher-level justification and higher-level knowledge, respectively.

What is needed, on such a view, is that the justified believer have "direct and privileged access" to what satisfies the conditions for justification, and the knower must have such access to what satisfies the conditions for knowledge. If Descartes thought that in order to know, he had to prove that whatever is clearly and distinctly perceived is true, then he would be an internalist of this type.

A weaker version of this kind of internalism would only require that one could potentially have the higher-level justification or knowledge. In this case, there need only be "available" to a someone with a justified belief "a method for discovering what the nature of justification is." What both conditions have in common is that they require for knowledge or justified belief "having epistemic access to the conditions for knowledge and justified belief."

The strong condition seems to be too strong. This is to say that the very definition of knowledge and justified belief seems at least to flirt with ruling out their possibility. The reason is that it seems to generate an infinite regress: in order to be justified in believing that p, one must be justified in believing that one is justified in believing that p, and so on.

The regress might be stopped by allowing that in some cases, there are beliefs that do not need higher-level justification in order to be justified themselves. Fumerton singles out non-inferential beliefs as candidates for stopping the regress. So even if inferential beliefs require higher-order justification to be justified, non-inferential beliefs do not. Fumerton himself accepts the principle that to be justified in inferring p from a body of evidence E, not only must E be justified, but one must also be "justified in believing that E makes epistemically probable p." This belief, however, may be non-inferentially justified.

Stopping the regress by appeal to non-inferentially justified beliefs is ruled out on the strong version of iterative internalism. BonJour's internalism is strong in this way. It is so strong, in fact, that it falls victim to the regress itself. We are said to get access to justification through coherence, but given strong iterative internalism, we need meta-beliefs about coherence in order to justify first-level beliefs. He really does not stop the regress by his device of presuming that such meta-beliefs are true, and Fumerton says that he practically admits to skepticism.

If strong iterative internalism is to be saved, it could only be through an infinite hierarchy of "increasingly complex intentional states" (beliefs about beliefs about beliefs, etc.). It may be possible to argue for such a view, by allowing perhaps that the needed states not be actual but only dispositional, and that nothing new is needed to justify belief at each higher level. But do internalists really want to "be painted into a corner this tight where the only escape is a view that might not even be intelligible?"

Even the weaker view seems to threaten to be psychologically implausible. Granting the higher levels of justification need merely to be potentially accessible, they must be at least understandable. And Fumerton expresses pessimism about his own ability to form coherent thoughts of high-level relations of thoughts. He concludes by noting that iterative internalism forces his view over to the side of externalism, but to do so would not be getting to the heart of the real debate between internalists and externalists.

To drive home the point that "access" is not the main bone of contention, he shows how an externalist like Goldman could accept weak iterative internalism. Since the access need only be potential, Goldman could say that one could have "access" to the fact that a belief-forming process is reliable. This would mean that there is a reliable process for determing whether the process is reliable. This might not be a problem if the processes might end up being the same at some level. And besides, they do not have to be actually used, so "it is not clear that the regress could be vicious."

It might even be that a reliabilist could indulge in strong iterative internalism. If we take the mind of be of sufficient (infinite) complexity, there might be an infinite hierarchy of reliable processes generating belief. The point of suggesting this is not to suppose that any reliabilist would actually adopt this view. It is only that a true-blooded internalist would say that this does not satisfy what the internalist is really after. So we must look elsewhere for the difference.

A less clear way of understanding the difference is in terms of "normativity." It is claimed that externalist accounts of knowledge and justified belief are not normative in the way that externalist accounts are. One way to understand this is that reliability or unreliability (for example) are not the basis for epistemic praise or blame. (Recall that Goldman had put these matters under the heading of "weak justification.") A demon-victim who is unreliable cannot reasonably be blamed for his false beliefs. Because of this, we cannot say that the demon-victim's beliefs are irrational.

Fumerton wants to distinguish carefully between charges of irrationality and charges of epistemic blameworthiness. To call a belief irrational is to criticize it. We do criticize a belief when calling it unreliable, but this is akin to criticizing a knife for being dull. The reliabilist would say the criticism of unreliable beliefs is that they do not lead us to what we want, which is true beliefs. For the same reason, we criticize a knife because it does not do what we want it to do, which is to cut things. So there is something of a "normative" element in externalism as well.

If we really want to make a distinction between internalism and externalism on normative grounds, it would have to be for deeper reasons. That is, an externalist account of justification and knowledge would have to be "deontological." They would have to make justification by definition a matter of being blameless, as Chisholm once did.

As an internalist, Fumerton does not wish to be saddled with this kind of definition. In the first place, there is no connection between epistemic worth and moral worth. "To criticize a person's belief is not to suggest that the person is morally reprehensible for having that belief."

Further, there is a problem of "conflicting duties." One has an epistemic duty to believe only what is true, while there is a practical duty to believe otherwise, say that you will get well, even with evidence that you will probably die. Fumerton suggests that a belief can be criticized even when a person is in absolutely no position to do anything about it and hence is morally blameless. "He may be doing the very best he can with the potential he has, even though his best effort still results in the having of irrational beliefs."

So here is another failed attempt to draw the line between internalism and externalism. Fumerton does not think that externalism properly handles the issue of the criticism of belief for being irrational. But at the other extreme, he thinks that viewing justification and knowledge deontically involves the same mistake as externalism commits, which will be discussed below.

Fumerton traces the origin of externalism to the problem of giving an analysis of epistemic probability. This is an all-important concept in any fallibilist account of knowledge. We want to say that one's evidence makes probable other propositions or beliefs which it does not logically imply. If the evidence does make a belief probable, to the correct extent, then we might say that the belief is justified. Since most of our beliefs that we think are justified rest on probabilities, whether or not we know depends very heavily on whether our evidence really does make our beliefs probable.

In the eighteenth century, David Hume in An Enquiry concerning Human Understanding advanced an argument that seemed to be devastating for the prospects of epistemic probability. Suppose one held that there is a principle to the effect that if many observations associate A with B, and there are no observations of A without B, then it is probable that all As are Bs. Here E is the observational evidence and p is the proposition that all As are Bs.

Hume argued that the claim that evidence E makes proposition p probable cannot itself be justified except by circular reasoning. The evidence applies to what has not been observed only if nature is uniform. Our evidence that nature is uniform can only be that nature has always been uniform. Thus we have evidence E* whose conclusion p* would be justified, if at all, on the basis of probability principle.

One way of responding to Hume's objection is to take our beliefs about probabilities to be foundational. Hume himself had taken such beliefs to be based on other beliefs concerning the frequency of observed events. On Fumerton's account, I can be directly acquainted with the fact that E makes p probable, just as I can be said to be directly acquainted with the fact that I am feeling an itch. Fumerton calls probability a "sui generis" epistemic concept, a concept "of its own kind."

The real basis of externalism, he goes on, is the attempt to give a certain kind of analysis of epistemic concepts, especially of probability. For example, after first favoring a direct-acquaintance view in The Problems of PhilosophyBertrand Russell tried to understand epistemic probability in terms of the comparative frequency of events (Human Knowledge: Its Scope and Limits). Roughly the frequency account is that, "one proposition E makes probable another proposition P when the pair is of a kind e/p such that usually when a proposition of the first kind is true, a proposition of the second kind is true."

We generally think that what underlies the e/p association is some law of nature, and so the frequency understanding of epistemic probability can be called "nomological" (from the Greek nomos: rule or law). Now let us take reliability as an account of epistemic justification. What makes one justified in a belief is that it was produced in a reliable way. To say that it is produced in a reliable way is just to say that there is a high frequency of correlation between the production of the belief and the truth of the belief. This is what makes the belief epistemically probable.

So, underneath the reliability account of justification is a frequency account of epistemic probability. Given that the frequencies are accounted for on the basis of causal laws, reliabilism reduces epistemic probability to nomological connections. The reliabilist, in making this reduction, has, in Fumerton's view, "defined away" "the concepts fundamental to his discipline; he is trying to analyze fundamental epistemic concepts in terms of other non-epistemic concepts." To put it another way, the concept of epistemic probablity has become "naturalized."

Is there any harm in the naturalization of epistemic concepts? Fumerton thinks that it is an attempt to "define the undefinable." (This is the problem for deontic internalism, which was alluded to earlier.) In the early twentieth century, G. E. Moore had made the same criticism of the attempt to naturalize ethical concepts such as "good" and "right." He had said that to attempt to do so is to commit what he called "the naturalistic fallacy." Fumerton disagrees that there is any naturalistic fallacy in ethics, but he says that there is one in epistemology.

The internalist, as described by Fumerton, raises philosophical questions about what the facts about epistemic probability are. What evidence makes what propositions probable, and to what extent? If we take the externalist route, then this is a question about causal connections and the frequencies they support.

But can we not investigate these? That would not be enough, according to Fumerton. This is because we want to be "acquainted" with these facts, to have them "before our consciousness." He goes on to claim that "Acquaintance with evidential connections would be impossible if evidential connections are to be understood in terms of frequencies or other nomological connections."

The internalist solution proposed by Fumerton is to hold that we are acquainted with epistemic probabilities, but in a direct, primitive, irreducible, undefinable way. The analogy, due to the economist-philosopher John Maynard Keynes (A Treatise on Probability), is with the way in which we are acquainted with the facts of logical implication or entailment. We know in a primitive, first-hand, way which premises imply which conclusions. Another feature of Fumerton's view is that we are directly acquainted with what the relation of direct acquaintance is.

Why do externalists look to causal connections to underwrite epistemic probability? Fumerton suggests that the reason is that they think that internalism makes skepticism unavoidable. They do not think that we can find direct acquaintance with epistemic probabilities. If we cannot, then we would have to find another source in consciousness of what makes a proposition epistemically probable given our evidence.

But Hume held that there is no such source. If he is right, then all of our beliefs that depend on epistemic probabilities are not justified. On Fumerton's analysis, externalism recognizes that Hume is right about the lack of justification for beliefs about epistemic probabilities, but "they do not want its truth to cheat us out of knowledge and justified belief." If you reduce epistemic probabilities to some facts about nature, then knowledge and justified belief are saved. "As long as nature (we now prefer to talk about evolution) has ensured that we respond to certain stimuli with correct representations of the world, we will know and have justified belief." Indeed, we can even extend knowledge to non-human animals and machines, as externalists have done.

The problem with this approach, as Fumerton sees it, is that it puts the cart before the horse. That is, it uses the intended result, to uphold "commonsense intuitions" about what we believe, as the basis for its accounts of epistemic concepts. The method of the externalists is to use whatever it takes to understand epistemic concepts so that they yield knowledge.

Fumerton acknowledges that there are senses of "knowledge" which this captures. But it does not capture "the philosophically relevant epistemic concepts." Which are these? Those whose satisfaction "resolves philosophical curiosity and doubt." As he has claimed, the externalist accounts do not do this, because we do not have acquaintance with the fact of epistemic probability.

He thinks we do have acquaintance with the relevant epistemic facts, but he is willing to acknowledge that we might not. "Hume may have been right--it may not be possible to justify in a philosophically satisfying way much of what we unreflectively believe." If so, then we must admit failure, rather than pretend that we have satisfied our philosophical cravings with an externalist account of epistemic concepts.