Philosophy, Film, Politics, Etc.

Monday, June 2, 2008

Omniscience, Testability and The Knowledge Argument

I am extremely grateful to Professor Torin Alter, who specializes in the philosophy of mind. I emailed him my previous post on the knowledge argument, and he was kind enough to reply with some guidance and feedback.

He informed me that other philosophers (including himself) have questioned P2. He also directed me towards avenues for further research, including his entry for The Internet Encyclopedia of Philosophy: "The Knowledge Argument Against Physicalism." Then he showed even more generosity by engaging my ideas. Specifically, he corrected my error in regarding P2 as a stipulation, as opposed to an assumption (I have since corrected the mistake); and he expressed strong doubts about my claim regarding the inconceivability of omniscience. He suggested that the notion of omniscience might in fact be a valuable guiding principle for science.

What follows is my response to Professor Alter. It is a defense of my views on omniscience and a more elaborate criticism of the knowledge argument.

Dear Professor Alter,

I apologize for the length of this email. I would not send it if I did not think it contained compelling and original arguments which will be of interest to you.

In defense of my claim about omniscience, consider that a mental state definable as "omniscient" would have to include all of the information in the universe. Such a mind would have to be as vast as the universe itself. Mary's televised lectures would have to cover every fact of the universe, and so could never be completed within the span of the universe. Mary simply cannot be omniscient.

Furthermore, omniscience cannot be scientifically defined, because there is no way to define omniscience in testable terms. (Testing for omniscience would require testing for every event in the universe, and would thus require the universe to end.) I would go even further, and say that all concepts are defined in terms of tests, so that omniscience can never be a well-defined concept in any domain. Thus, the notion of omniscience is not only antithetical to science; it is simply unthinkable. We can only speak about the potential for omniscience, as the ability to perpetually learn more and more facts, and not actual omniscience.

Perhaps a "final theory" is possible, in which all of the laws of nature are contained within a finite set of formulae. However, such formulae could not contain all of the information in the universe. They would instead contain the information required to provide tools which, when coupled with new information, would allow us to make the most accurate predictions possible. Even the most perfect scientific theory will exclude an incalculable amount of information, because we require that extra information to implement and test the theory.

Testability is the key here. In my view, physicalism is definable in terms of testability. That is what science is all about. We need not limit our notion of "the physical" to whatever our current scientific models describe. And we need not imagine a perfect, completed science, either. We need simply regard science (and all learning) as an ongoing process, and not a static set of facts. What is "physical" is that which we can define and repeatedly test for. In short, it is everything.

I know that rejecting the knowledge argument by attacking the premise of omniscience is not wholly satisfying. It seems like physicalism is getting off on a technicality. My preceding argument may be sufficient to defeat the knowledge argument, but it does not address the intuition which tells us that, no matter what Mary learns inside that black and white room, she will always learn something new about the experience of color when she goes outside and sees colors for the first time. It should not matter if she learns "all of the facts" inside the room or not.

We may therefore formulate a different version of the knowledge argument, one in which the issue of omniscience is completely avoided. In this version, Mary does not learn "all of the facts" about color vision; instead, she learns any arbitrarily large set of facts about color vision. In this way, we only need to claim that, no matter how much Mary learns inside the room, she will always learn something new about the experience of color when she leaves the room.

There will always be some information about color vision missing from Mary's televised lectures. We should not be surprised, then, that Mary will always learn new facts about color vision upon leaving the room. Yet, why would Mary never learn what it is like to see colors? If Mary can learn an arbitrarily large set of facts about color vision, why is she never able to learn about the experience, the phenomenal character, of color vision?

To answer this question, I turn to the Ability Hypothesis, which (in my slight reformulation) I think is robust enough to defeat both the original knowledge argument and my revised version.

In your paper, "A Limited Defense Of The Knowledge Argument," you criticize the Ability Hypothesis by saying that "those sorts of abilities [required for know-how] would also seem to be required for retaining garden-variety factual knowledge, e.g., knowledge of historical dates. This does not tempt us to reduce garden-variety factual knowledge to the possession of abilities" (p. 38).

I disagree. I think we can and must reduce all forms of knowledge to the possession of abilities. However, I wish to part from the traditional Ability Hypothesis by claiming that know-how is informational; however, it is non-propositional. All knowledge is information, and all information is a matter of ability. (After all, what distinguishes information from noise is its ability to inform.)

The set of abilities we associate with speech are complex and unique, and deserve a special place in our lexicon. Yet, we should not forget that they are abilities. This is apparent when we consider how it is that we recognize propositional knowledge as such. We require some sort of test. Propositional knowledge is definable only in so far as it can be tested for. Whenever we test for something, we are testing for the ability to pass that test. (I appreciate Dewey's comment: "It should be possible to discern and describe a knowing as one identifies any object, concern, or event. It must have its own marks; it must offer characteristic features—as much as a thunder-storm, the constitution of a State, or a leopard.")

In another paper, "Know-how, Ability, and the Ability Hypothesis," you reject the Ability Hypothesis on the grounds that know-how does not reduce to ability. You write, "A guitar teacher might know how to play well, but lack the ability to play well. Perhaps he's uncoordinated or never practices. Further, his ability to play might improve with practice, even as his know-how remains constant. So, his know-how doesn't reduce to the corresponding ability" (p. 5).

If the guitar teacher does not practice and cannot be relied upon to play well, then in what sense does he know how to play well? I propose that he does not know how to play well. Maybe he once knew how, and maybe he has some recollections about that former ability; but such a state is not equivalent to knowing how to play well. It is, at best, knowing some things about how to play well.

I find the example of Juan, Chomsky's hypothetical aphasia victim, equally problematic. To quote your paper again, "As Chomsky notes, to maintain that Juan retains his ability to speak and understand Spanish before the effects of his injury recede, the objector must invent a new concept of ability, designed to have the typical properties of knowledge" (p. 6).

What is required is not a new concept of ability, but a clearer understanding of how the knowledge in question involves particular abilities. The distinction between knowing-how and knowing-that must be defined, not as a difference between information and non-informational abilities, but as a difference between types of abilities.

Propositional knowledge is unique in that it makes abstract predictions. For example, the proposition "this fruit is an orange" makes predictions about how the terms "fruit" and "orange" can be applied to particular experiences. There need not be a clearly defined limit to the number of possible tests described by a proposition. What must be clear, however, is that some conditions are specified which would allow us to regard the proposition as either true or false. When we say a person has propositional knowledge, we mean that they have the ability to properly make a set of abstract predictions.

Propositional knowledge relies on a form of know-how, but one which involves a particular sort of experience; namely, the experience of language and abstract thought. All propositional knowledge can be communicated in any medium, because the information contained in any language is not dependent on the medium in which that language is ever expressed.

Know-how (apart from propositional knowledge) is not about abstraction. The abilities we mean when we talk about know-how, as opposed to knowledge-that, relate directly to other forms of behavior. It is for this reason that a person can know how to tie their shoes, or know how to play billiards exceptionally well, or even know how to speak eloquently about the American Civil War, but be unable to describe the process effectively.

Mary does not learn new propositional knowledge when she leaves the room. Rather, she gains new, non-propositional abilities. This view is usually said to be contrary to the position maintained by supporters of the knowledge argument. However, as it turns out, the opposite is the case: Supporters of the knowledge argument must agree with the Ability Hypothesis (as I have formulated it). The reason is thus: if they were to claim that Mary's new knowledge is propositional in nature, then they could not claim that this propositional knowledge had to have been absent from the televised lectures. For, again, propositional knowledge can be transmitted in any medium.

If we are going to maintain that Mary inevitably learns something new about the phenomenal character of color vision when she steps outside of the room, then we cannot maintain that this new knowledge is propositional knowledge. It must be some other set of abilities, or know-how.

Thus, supporters of the knowledge argument must accept that Mary's newfound knowledge is know-how. The question they press is whether or not this know-how is physical. Yet, as I think is obvious, there is no sense in claiming that it is not physical. For, if know-how were not physical, it could not be used to test propositional knowledge. The very fact that we use our experiences of color to make tests about the physical world is incontrovertible evidence that our know-how is part of that physical world. There is no grounds for introducing any non-testable elements into the equation. To claim otherwise is to deny the requirements of sense.

To put this all in a clearer perspective, consider the idea that, before she leaves the room, Mary doesn't have any knowledge of color vision at all. Her propositional knowledge is, at best, mere belief.

Mary has gained a great deal of propositional knowledge about color vision, but is unable to apply those propositions to new experiences so long as she remains in her black and white world. She cannot test her knowledge to determine if it is legitimate. For all she knows, the real world of color vision is wholly unlike everything she has learned. Therefore, technically speaking, Mary does not have knowledge of color vision, even though she can recite facts about it to an arbitrary degree of accuracy and completion. Her beliefs are true, but they are not justified, because she cannot test them.

Say we give Mary a telephone and fax machine so she can correspond with researchers in the world of color, guiding their studies with her extensive knowledge base. This system (we can call it the Mary-System) has color vision, because the researchers are using equipment to test Mary's predictions. Yet, Mary remains confined to her black and white world. In this scenario, a proper knowledge of color vision has been reached; however, that knowledge does not reside wholly within Mary. It resides within the Mary-System. (This scenario is reminiscent of Searle's Chinese Room, where the system understands Chinese, even if the man inside does not.)

Upon leaving the room, Mary loses the predictive power contained in the Mary-System. Yet, she gains the ability to test her knowledge of color vision directly, on her own. She may never improve upon the theoretical framework she learned inside the room, but she has finally been able to apply that framework directly to cases of color vision. Thus, her abilities are expanded (she gains new knowledge) by virtue of her own, previously unused neurological color-detecting devices. Yet, any propositional knowledge so acquired could easily have been available to her inside the room. The only difference is that now she can test the knowledge directly.

I think the above stands on its own as an argument for my views. I will now end by addressing one of your arguments against the Ability Hypothesis.

You write, "In short, one might retain knowledge of how to speak and understand a language even while lacking the ability to do so; therefore, knowing how to speak and understand a language can't be identified with the corresponding ability."

To see more clearly why I do not accept this point, consider a variation of Searle's Chinese Room. The man inside the room operates a complex series of levers which, when coupled with his elaborate instruction manual, allow the room to produce outputs that convince every outsider that there is a real understanding of Chinese going on. The man inside the room does not understand Chinese, but the system as a whole creates the impression of such an understanding. The system understands Chinese.

Now, there is a malfunction and some of the levers stop working. The system has broken down. The man still does not understand Chinese; yet, now the system does not understand Chinese either. The ability to communicate in Chinese has been lost, and no one element of the system can be said to know how to communicate in Chinese. After a little while, the man fixes the levers, and understanding is restored to the system.

If any part of the system had been able to demonstrate an understanding of Chinese, then we would not have been able to say that the knowledge was lost. And if any manifestation of the knowledge had occurred, we would not have been able to say that the ability was absent. This demonstrates that "the knowledge of how to speak and understand a language" is inseparable from "the ability to do so."

My argument here differs slightly from the access argument you address in your paper. The question is not about access, but functionality. In the Chinese Room, there is a complete set of instructions for operating the system. These instructions may even be internalized by the man inside. His knowledge is not lost, and access to his knowledge is not lost; but his knowledge was never sufficient for an understanding of Chinese. If it had been, then we would have said that he understood Chinese.

Thank you for your time. I eagerly yet patiently await your response.