Dec 3: Reliabilism and naturalized epistemology
|
|
|
To do for next class - 1) Read chps. 8 & 9; 2) Give specific examples to explain the generality problem; 3) Precisely what is it that an epistemic nonnaturalist asserts?. |
To do for next class - 1) Read chp. 10; 2) Be a skeptic; 3) Final papers due. |
Introduction to reliabilism
The three main proponents of reliabilism in epistemology has been Alvin Goldman, Fred Dretske, and John Nocizk. The intuition behind his position is that, in order for a belief to be justified, it must be produced reliably. As an analogy, David Armstrong compares the reliability of belief forming processes to the reliability of a thermometer indicating the temperature of a room. He believes that we should find the same kind of nomic (law-like) relation between beliefs (the thermometer) and truths (the temperature). Goldman suggests a number of processes which he believes to be reliable ones. Do you recall what they were? (Perception, remembering, good reasoning, introspection). What were these contrasted with? (a hunch, wishful thinking, bad reasoning). Do you think this would be considered and internalist or externalist theory of knowledge? (externalist, the justification is internal, but truth determination is external). Being careful to account for the possibility of evidential defeat, we can formulate reliabilism as:
Someone's belief is justified if and only if 1. it is produced by a reliable cognitive process; 2. there is no alternative reliable process that, added being used, would have caused that persons belief not to be formed.
Steup identifies three problems for process reliabilism. He also briefly touches on a fourth, to which we will return, which is that though purported to be a naturalistic theory, reliabilism may not escape using evidential is language in analyzing epistemic justification.
The evil demon problem
Do you recall what this problem is? (Supposed to demon makes all of your beliefs seem like they are about real objects when in fact they are not. Then your perceptual beliefs are all false and hence unreliable). Given this scenario, our intuitions tell us that our perceptual beliefs are not reliable, if they are justified since they are phenomenologically indistinguishable from "real" perceptual beliefs. Of course, this sort of scenario can be run for any process we would otherwise being reliable. How, or why, good a process reliable is avoid this difficulty? (There is a deep divide between internalist and externalist theories which is evident in this problem. Externalists will always be subject to these criticisms, but claim that they are unreasonable situations and probably don't deserve our attention. In other words, they would employ Occam's razor to argue against the validity of such concerns).
How does Goldman attempt to avoid this problem? In at least three ways. First, he introduces another formulation of reliabilism which he calls rule reliabilism. This is very similar to process reliabilism in his formulation, but introduces the idea of "normal world chauvinism". This allows for reliabilism to be determined with respect to "normal world" which automatically rules out evil demon worlds. Thus, the problem assault but it introduces another difficulty. Do you recall what it was? (It makes beliefs produced by reliable processes not present in the normal world unjustified, which seems implausible).
As another means of approaching the evil demon problem, Goldman introduces the distinction between strong and weak justification. Do you recall what that distinction is? (Strong justification: de facto justification through reliable processes. Weak justification: justification derived from epistemic blamelessness, though unreliably produced.). Goldman provides the example of the community dependent on astrology. This distinction, then, insinuates that evil demon world people only have weakly justified beliefs (because perceptual processes are, in fact, not reliable). What would a proponent of internalism say about this solution? They would deny that there is anything weak about the justification enjoyed in the evil demon world. Rather, they have as much justification as we, in the real world, have. In other words, they would claim that reliability of production has nothing to do with justification.
Goldman's third solution to the evil demon problem introduces virtue reliabilism. Not surprisingly, this theory claims that a belief is justified if it results from intellectual virtue and unjustified if it results from intellectual vice. Goldman claims that we inherit a list of cognitive virtues from our social background. When evaluating the claim, we compared to this list candidate one of three statuses. It is either justified (matches virtues), unjustified (matches completely or partly vices), or non-justified (is not our list). What, most obviously, does the inclusion of this "list" in this theory do for Goldman? (It seems to solve the generality problem, since generality is specified by the list). How does this version of reliabilism to solve the clairvoyance problem? (In the case when the clairvoyant ignores contrary evidence we've assume that would be our list of vices; in the case where we have no other evidence, the belief is unjustified if clairvoyance is a vice or it is non-justified otherwise). How would this solve the evil demon problem? Golden would agree that in the evil demon world the victim has justified beliefs. Why? (Because, like in our world, perceptual beliefs to be a list of virtues). However, we are using a different notion of reliability here. What are the two notions? (De facto reliability; vs. deemed reliability). This difference is an extremely important one. For one, it makes justification depend on the evaluator of a situation (we would judge the evil demon beliefs unjustified whereas a victim would judge them justified on this theory: note that Steup thinks that this is an undesirable result do you agree?). As well, it internalized is the previously external criterion of reliability. Does this mean we have a internalist theory now? (An internalist theory of justification, but not of knowledge). What you think Goldman's response would be to the problem of having mass madness which permitted unreliable processes on the list? (He would note that he is doing descriptive epistemology that is trying to predict which beliefs will be deemed justified in which unjustified by our current practice. This is unsatisfactory, though, for classical epistemology.).
What is this example trying to take away from reliability theories? (Their necessity, i.e. this show is that reliability is not necessary for justification).
The clairvoyance problem
What is the clairvoyance problem? The problem of Norman knowing where the president is at any moment. This is very similar to the neurophysiologist telling Spencer that his childhood memories are unreliable. In both cases, reliable processes (clairvoyance and memory) do not immediately produced justified beliefs. Thus, reliable belief production is not sufficient for justification. However, this is not the claim in process reliabilism as stated above. Rather it is that undefeated reliable belief production is sufficient for justification. If the original counter example had succeeded what would have shown about reliabilism? (That it is not sufficient for justification)
However, Steup claims that Goldman runs into a problem in describing what counts as undefeated reliable beliefs. Do you recall what that problem was? It has to do with claims to naturalizing epistemology by process reliabilists. In particular, the reliabilist must introduce processes such as "taking into account contrary evidence" or "taking into account the absence of supporting evidence" in order to handle cases like the clairvoyance one. Steup seems skeptical that these kinds of processes can be given the naturalized descriptions. Only in the face of this uncertainty are such processes a liability for this theory.
The generality problem
Before discussing the problem, do you recall what the differences between a token and a type? (A token is an instance of a type). So, can you give an example of a token process of the perceptual experience type? Is that token reliable? It is a category mistake to apply the notion of reliability to tokens. They determine the reliability of a type but reliability itself is only a property of types of processes. However, there's nothing to respect to the number of tokens which determine the reliability of a type. Therefore, Steup introduces the problem of OTOP's (one token only process types).
What is an example of an OTOP? And how is it a problem for reliabilism? (The process type that the token specifies is perfectly reliable only because there is a single instance). In this case, the type is too specific. Can you give an example in which case the type is too general? (Vision as a reliable process; the example of the faraway object). Can you give an example of how a belief would be unjustified whenever it is false under simple reliabilism? (Form an OTOP for an incorrect perception, in which case that OTOP is perfectly unreliable, can always be formed, then implies that the belief is unjustified). This, then, is the generality problem. How do we specify processes of the right amount of generality that is, neither too specific nor too general.
Naturalized epistemology
Simply put, to naturalize any subject matter is to make it responsible to empirical science. That is, to use the methods and criteria for good science in evaluating theories forwarded about that subject matter. Historically, epistemology has not been naturalized. Recall what Descartes method and conclusions were. To summarize, Descartes employees rationalism and solipsism. Rationalism because he relies on reason as a source of knowledge (that he cannot be mistaken in believing he is thinking) and solipsism in his allowing himself to doubt everything but his own mental states (I am thinking). Steup recounts Chisholm's criteria for doing naturalized epistemology. Do you recall what they were? (1. There is something we know and it is what we expect. 2. We can know what it is a constitutes our ground or reason or evidence for thinking that we know something. 3. There are valid general principles which state the conditions under which we have reasons for what we believe.). How does this compare to Descartes position? (Chisholm rejects solipsism but accepts rationalism).
What does it mean to say Chisholm takes the questions of epistemology to the Socratic questions? (A Socratic question is one which we can answer by drawing from the body of evidence and knowledge we currently have, whenever the question is asked.). So, Chisholm leaves we can discover the general principles of epistemology and do so from the armchair. In other words, traditional epistemology, like that of Chisholm and Descartes is constituted by the claims that (1) epistemological questions are Socratic band; (2) answers to those questions are a priori knowable necessary truths. Agreement with these two principles means we do not need to engage in any sociological are psychological research to understand epistemology. Rather, we, by being human beings, are in a position to answer the questions of epistemology no matter where we may be as long as we can think, reflect, intuit, and meditate.
W. V. O. Quine introduce the notion of naturalizing epistemology in the paper called "epistemology naturalized". He does so by both showing that traditional epistemology has failed and arguing that natural science can succeed in its place. Quine thinks of traditional epistemology as being a doctrinal project concerned with identifying the foundation and performing deductions from the foundation to beliefs about the physical world. Quine also identifies the conceptual project (one followed to a large degree by Carnap and philosophers known as the logical positivists) which is concerned with providing definitions for translating talk about physical bodies into talk about sense impressions. Not only did Carnap fail, Quine thinks that it is in principle impossible to succeed (the introduces a thought experiment about translation of the word "gavagai" to "rabbit"). So Quine asserts that the quest for Cartesian certainty will not succeed. The alternative he provides is to study the relation between our sense impressions and are theories about the world. He feels we should reject a concerned with justification or quest for certainty rather we must study scientifically (or psychologically) what happens in our brains, i.e. the natural phenomenon in a human subject. In essence, he has removed normativity from epistemology. What kind of defenses do Steup suggest against Quine's argument? ((1) there is an artificial coupling of psychology with the rejection of normative epistemology; (2) arguing that there is more than Cartesian foundationalism to normative epistemology. Quine is, of course, aware of these other forms of epistemology but feels they cannot justify their normative aspects).
Epistemological naturalism
Another way of understanding the project of naturalizing epistemology can be provided by paraphrasing Gilbert Harmon when he notes that a naturalist wants to be able to locate justification, truths, and belief, etc. in the world the same way that tables, colors, jeans, temperatures and so on can be located in the world. What epistemological naturalists are doing is to make claims about the ontological makeup of the world and provide an analysis of how that ontology undergirds epistemic values. What concept does this position remind you of? (Supervenience) the nonnaturalist would claim that epistemic facts are brute fundamental facts, not ones related to nonnormative descriptive facts. Naturalists, then, have the unique project of describing the way in which epistemic facts are related to descriptive facts. This then is a restatement of the supervenience thesis. So again, we must as naturalized epistemologists identify which descriptive properties are such that, necessarily, any belief with those properties is justified. Of course, naturalism expects criteriological analyses (that is analyses of evaluative concepts in terms of nonevaluative ones).
So, we can identify after Steup two different projects an analytic and a metaphysical one. The first is a task of specifying in nonnormative terms on which nonnormative properties epistemic justification supervenes. The second is to argue for the claim that every beliefs epistemic status supervenes on its descriptive properties.
Chisholm vs. Goldman
Chisholm is supposed to be the critic of naturalism. However, he endorses both the analytic and metaphysical projects. Goldman, the proponent of naturalism endorses same projects. How then do they differ? What are the three stages or states that Goldman identifies in constructing epistemological theory? ((1) forming a by conditional between belief justification and the right system of justification rules. (2) specify a criterion for rightness of the rules. (3) provide a detailed description of the contents of that rule system). What role is there for cognitive science (i.e. the naturalistic aspect) in this epistemological project? (In stage 3.). Up until this stage, the armchair is the place for both epistemologists (in this sense Steup refers to Goldman's position as conservative). However, in the third stage it is cognitive science that will do the work of determining the content of the theory for Goldman. So, unlike Quine's radical naturalizing of epistemology where the armchair is destroyed, Goldman's conservative project is to begin with a priori considerations in the armchair and eventually move to the lab. In contrast, though Chisholm accepts the aims of epistemology as conceived by the naturalist (what is that? To determine which nonnormative properties epistemic justification supervenes upon), he does not care, in particular, Ford cognitive science since the content of his theory does not depend on a scientific understanding of perception.
Though they share the same analytic objective, Goldman and Chisholm differ in their approach. Chisholm wishes to determine the set of sufficient nonnormative conditions for justification without identifying a necessary property while Goldman desires to determine both what is sufficient and necessary for justification (in his case reliability). Both of these positions are susceptible to the skeptical claim that we cannot possibly identify the conditions sufficient for justification without using epistemic terms. Do you recall before problems associated with Goldman's project? (1) the problem of Infinity: there might be an infinite number of nonnormative justification in sailing base properties in which case Goldman's project of identifying a necessary condition will fail though Chisholm is less ambitious project could still succeed. (2) the problem of magnitude: the base properties might be finite but very large in which case Goldman's project is impractical though possible; in which case Chisholm is on the right track perhaps (3) the problem of unity: if there are a small number of base properties then the disjunction that we form as the analysans of "a belief is justified if it only if it exemplifies..." is uninformative. In other words we're not guaranteed finding a core property. (4) the problem of certitude: how would we ever determine that we have the right finite subset of base properties? What kinds of problems does the pessimists note for Chisholm's project? (1) the problem of complexity: it is really hard to come up with a nonnormative property or conjunction of such properties that is sufficient without introducing normative concepts like defeasibility.
Naturalism, pessimism, and antiapriorism
What is the pessimists claim amount to in a naturalistic context? It announced in the claim that we know that epistemic justification supervenes on nonnormative base properties but we can't say how it supervenes. In other words, whenever beliefs different epistemic status we can agree that there must be a nonnormative difference between the however these differences don't seem to display a law like regularity, that is a regularity that we can identify necessary and sufficient conditions for. The pessimists can, however, provide the analysis like the one Steup notes i.e. by simply stating that perception helps create justified beliefs but does not invariably result in justified beliefs, in fact nothing does.
What does antiapriorism deny? It denies that's a priori epistemology helps determine anything about epistemic justification (unlike Goldman the thinks that it can help determine what epistemic justification supervenes on). In general, philosophers like Kornblith and Maffie argue that a priori methods are bad ways of coming to know something. This is an extremely difficult point to argue very broadly. Recall the argument that the skeptic can never prove to the non-skeptic that a priori skepticism is true since the skeptic would have to rely on an understanding of what constitutes valid arguments. Thus, the epistemological dualism (knowing about values in one way (a priori) and facts in another way) is acceptable.
The Darwinian argument
This argument runs as follows: since we are a successfully propagated species then we are likely to have mostly true beliefs since if we didn't we would not be able to successfully reproduced. In this sense, we have a "bias in favor of true beliefs". This argument to summarize on p. 198. Steup notes two problems. The first is accepting that processes biased towards truths are the ones by which we ought to form are beliefs. Would we know this a priori or directly? If empirically what would be valid evidence for this? The second is trying to determine how biased we really are, and more importantly which ones are properly biased. Since we know that some are not, like wishful thinking, we can really only conclude that some processes by which we form are beliefs are the ones by which we ought to form are beliefs on the basis of this argument.
So, this particular form of the Darwinian argument fails. However, as Steup notes there is a bigger picture to be seen here. It is a particular conception of naturalizing epistemology which is to answer normative questions through empirical means, that is rejecting a priority.
To do for next class
1) Read chp. 10; 2) Be a skeptic; 3) Final papers due.