Wednesday, April 28, 2010

Defending the Incompatibility Argument

In my last post, I argued that the knowledge argument and the conceivability argument are incompatible, which is interesting if only because they are commonly supported together in attacks against physicalism. Torin Alter has just emailed me saying that Mary Z has appeared in the literature already, and even in an argument similar to--though perhaps not identical with--my own (see McGeer, 2003). I haven't read his references, so I cannot comment on any of that yet. What I want to do here is defend my incompatibility argument against possible objections, including one raised in Torin Alter's email.

First, a brief review of the incompatibility argument:

  1. If zombies are conceivable, then we can conceive of Mary Z, a zombie version of the original Mary (hereafter 'Mary O').
  2. If zombies are conceivable, then Mary Z gains knowledge when she leaves her black-and-white room.
  3. If zombies are conceivable, then Mary Z's new knowledge does not entail non-physical facts.
  4. If zombies are conceivable, then Mary O gains whatever knowledge Mary Z gains.
  5. Therefore, if zombies are conceivable, then Mary O gains new knowledge which does not entail non-physical facts.
  6. If the knowledge argument is sound, then Mary O's new knowledge entails non-physical facts.
  7. Therefore, if zombies are conceivable, the knowledge argument is not sound.
  8. Therefore, the conceivability argument and the knowledge argument are incompatible.
Supporters of both the conceivability and zombie arguments might respond by claiming that Mary Z does not learn what Mary O learns: Mary O learns what it is like to see colors, but Mary Z cannot learn what it is like to see colors. However, this response is not to the point, because the incompatibility argument says nothing about whether or not Mary O learns anything over above what Mary Z would learn, if she were conceivable. The incompatibility argument is more subtle than that.

The incompatibility argument shows that, if zombies are conceivable, then Mary O gains new knowledge which does not entail non-physical facts. This is true regardless of whether or not Mary O also happens to gain knowledge which does entail non-physical facts. Therefore, the fact that Mary O gains new knowledge is insufficient to draw the conclusion that she gains knowledge which entails non-physical facts. If we want to claim that Mary O gains knowledge of non-physical facts--or knowledge which entails non-physical facts--then we must claim more than what is stipulated by the knowledge argument. Specifically, we must stipulate that phenomenal knowledge entails non-physical facts; but that is precisely what the knowledge argument was supposed to demonstrate. We cannot reformulate the knowledge argument with that stipulation without begging the question against physicalism.

Torin Alter objects: "I don't see why we should agree that zombie Mary learns anything. Of course, she may say that she has learned something, but why should we believe her?"

Of course, zombie enthusiasts will not grant that Mary Z learns what it is like to see colors. However, I think they must accept that she learns something upon her release from her black-and-white room. To see why, imagine the following experiment.

Before Mary Z leaves her black-and-white room, we show her some colors. She says they look unlike anything she has ever seen before. As Alter notes, we need not believe her. However, thanks to a machine scanning her neural activity, we see that her brain is in states it has never been in before. Something new is going on in her brain which correlates with her reports of new experiences, as the zombie argument requires--for Mary Z must be a physical and functional duplicate of Mary O. So Mary Z need not be experiencing what it is like to see colors (whatever that is), but something new is going on in her brain. Does this "something new" constitute new knowledge? Maybe not. Let's continue.

Now we teach Mary Z the names of some of these colors, and we test her memory of the names of the colors. Again, during this activity, our neural scan shows unprecedented activity. She seems to be able to do something she couldn't do before. In fact, her neurological activity is exactly like Mary O's when we performed the same experiment on her. So it would seem that, even if Mary O experiences something Mary Z has not experienced, they have both acquired new abilities which are in some respect the same. Does this new shared ability constitute new knowledge? I think it does. Mary Z learns to use names to correctly identify colors.

Some might argue that my proposed experiment would not work the way I claim. They might say that Mary Z could not learn how to use names to identify colors. Yet, if this is true, the conceivability argument fails; for zombies are supposed to function exactly like human beings. If human beings can identify colors with names, then so can zombies. The only difference, supposedly, is that normal human beings can do more--we can also access non-physical properties of color vision.

Indeed, if we suppose that zombies cannot learn to identify colors, then how could they learn to identify any objects? How could they identify tables or chairs? How could they identify sounds or textures? A zombie who could not learn how to identify objects and properties would not be able to learn anything at all, and would not function like a human being. Thus, for the conceivability argument to work, zombies must be able to learn how to use names to identify colors. And clearly Mary Z cannot do this from inside her black-and-white room. She must wait until she has access to colors (even if she lacks access to their phenomenal properties, if such properties be non-physical).

I can see only one way to make Alter's objection work, and it is not attractive. It requires regarding knowledge as absolutely unobservable, as absolutely disconnected from observable behavior. If we are to deny that Mary Z gains knowledge, then knowing cannot be demonstrated in any way, by any set of abilities or capacities. Thus, for Alter's objection to work, we must supplement both the zombie and knowledge arguments with an implausible view of knowledge--a view which holds that knowledge is not demonstrated with the correct use of language, or the learned following of rules. If this view is adopted, we must eternally wonder how anyone could ever know that anyone ever knew anything.


References


McGeer, Victoria (2003) "The Trouble With Mary" Pacific Philosophical Quarterly, 84, 4, pp. 384-393.

Friday, April 23, 2010

Are The Knowledge and Conceivability Arguments Compatible?

Two of the most popular and widely discussed arguments against physicalism are the knowledge argument and the conceivability argument. The knowledge argument is often described in terms of Mary, a super-scientist who learns every physical fact there is to know about color vision before she ever sees a color. She does this by reading black-and-white books and watching black-and-white science lectures from inside a black-and-white room. When she leaves the room and enters the polychromatic world, she experiences the color red for the first time. Only then does she learn what it is like to see red. Thus, the argument goes, Mary learns new facts about color vision. These new facts cannot be physical facts, because she had learned all of those already. So there must be facts about color vision which are not physical facts. Thus, physicalism is false--not all of the facts are physical facts.

One response to the knowledge argument is to reject the premise that anyone could ever learn all of the physical facts, if there even is such a thing as "all of the physical facts." I thus have offered a revised knowledge argument (see my last post and also this older one). In the revised version, Mary does not learn every physical fact. She only learns an arbitrarily large number of physical facts. Yet, no matter what she learns from inside her room, she will only learn what it is like to see red after she leaves the room and experiences red for the first time. Therefore, what she learns from seeing red cannot be a physical fact.

In my last post, I argued for the ability hypothesis, which claims that what Mary learns from seeing red is not factual knowledge. It is a more primitive sort of knowledge which is not propositional, and which cannot be learned by didactic discourse. I am not going to argue for the ability hypothesis again here. I am not going to critique the knowledge argument, either. I only want to demonstrate that it is incompatible with the conceivability argument.

The conceivability argument is commonly discussed in terms of zombies. Zombies, if they are possible at all, are beings which are physically and functionally identical to human beings, but which lack phenomenal consciousness. There is nothing it is like to be a zombie. Obviously, zombies cannot know what it is like to see red, because there is nothing it is like for a zombie to see red. This is not to say that zombies cannot see. Zombies are physically and functionally identical to human beings, so there is a sense in which they can see and even recall visual images. It's just that their vision does not have any phenomenal properties (whatever that means).

The zombie version of the conceivability argument goes like this (See David Chalmers, The Conscious Mind, 1996):

  1. In our world, human beings have phenomenal consciousness.
  2. We can conceive of a zombie world: a world identical to ours in every physical way, but in which there are no conscious beings.
  3. Therefore, the positive facts about consciousness in our world are not physical facts.
  4. Therefore, physicalism is false.
It is common to take issue with the second premise in Chalmers' argument. On the one hand, some claim that we cannot conceive of zombies at all. On the other hand, it is claimed that the second premise assumes that physicalism is false, and therefore cannot be used to argue for the conclusion that physicalism is false. I am sympathetic with both of these complaints. It does seem that any argument for premise 2 would be a direct argument against physicalism. And it does seem that zombies are inconceivable: I cannot imagine what phenomenal consciousness might be, if it could be absent from a being which was physically and functionally identical to a human being.

Obviously I am not sympathetic to the knowledge and conceivability arguments. Both should be rejected. However, instead of arguing directly against these two arguments, I want to pit them against each other. I want to suggest that these two arguments, though often supported together, are in fact incompatible. If I am right, then those who wish to reject physicalism cannot rely on both the knowledge and conceivability arguments. They have to choose.

Before I give my argument, I must present another thought experiment. Imagine the knowledge argument as before, but this time, Mary is a zombie. Zombie Mary learns all of the physical facts--or she learns any arbitrarily large number of physical facts--while inside her black-and-white room. Will Zombie Mary learn something new about color vision when she leaves her room and sees the color red for the first time?

According to Hugh Chandler (University of Illinois, Urbana-Champaign), who first brought up Zombie Mary at PhilPapers, she will. He writes: "She can now invoke an inner picture of red roses, can’t she? She can induce the necessary brain states. This is something she couldn’t do before" (Chandler, PhilPapers post 3539). The suggestion is that Mary will gain a new physico-functional knowledge of color vision. According to Hugh Chandler, this suggests that the difference between Mary and Zombie Mary is impossible to conceive.

While I give Hugh Chandler credit for introducing Zombie Mary to the discussion, he has not made the argument I am about to make here. Rather, his aim is only to show that Zombie Mary is not clearly conceivable. Now, I agree with Hugh that Zombie Mary is inconceivable. However, I don't think Zombie Mary gives us any new reasons to doubt the conceivability of zombies. A defender of Zombie Mary might respond to Hugh by saying that, while Zombie Mary does learn something--let's say she learns how to imagine, remember and recognize visual states of color vision--what she learns does not involve phenomenal consciousness. Zombie Mary can imagine red, for example, but when she does, there are no phenomenal properties of redness involved. So she does not gain the phenomenal knowledge which ordinary Mary gains. There is still nothing it is like to be Zombie Mary. While Mary gains phenomenal knowledge, Zombie Mary only gains a set of non-propositional abilities. So Hugh is wrong to think that a difference between Mary and Zombie Mary is inconceivable.

According to the ability hypothesis, possessing the abilities to remember, recognize, and imagine colors is identical to knowing what it is like to see red. If the ability hypotheses is correct, Mary and Zombie Mary both learn the same thing, and we have no basis for distinguishing between them. I think the burden is on Zombie Mary's supporters to argue for a distinction between phenomenal knowledge and these abilities, though others might say the burden is on supporters of the ability hypothesis to demonstrate that zombies are inconceivable. Either way, Zombie Mary has not helped us decide whether or not zombies are conceivable.

The knowledge argument is sometimes used to motivate the claim that zombies are conceivable. Chalmers (1996), for example, utilizes the knowledge argument to help motivate the second premise of his zombie argument. As I will now show, the knowledge argument cannot be used to support the conceivability argument. If Zombie Mary is conceivable, then the knowledge argument is invalid.

Recall how the knowledge argument goes. Mary, upon seeing colors for the first time, learns something new about color vision. She learns what it is like to see colors. Therefore, she could not have known these facts from inside her black-and-white room. But Zombie Mary also learns something when she sees colors for the first time, and what she learns cannot be a fact about phenomenal consciousness, because she never experiences that. Zombie Mary only learns physical facts, or gains physical capacities. Therefore, the fact that Mary learns something new upon seeing red for the first time is not sufficient to draw the conclusion that she learns non-physical facts about color vision. As I wrote in response to Hugh (where "Mary Z" refers to Zombie Mary, and "Mary O" refers to the original Mary):

"Our intuition tells us that Mary Z and Mary O both learn something new when they leave their respective black-and-white rooms. Presumably, whatever Mary Z learns, Mary O will learn as well. But then we cannot say that whatever Mary O learns is [non-physical] phenomenal knowledge, since Mary Z cannot gain [non-physical] phenomenal knowledge. Whatever Mary Z learns, it can supervene on the physical. We have thus lost the motivation to claim that Mary O learns something which does not supervene on the physical.

Thus, Mary Z, if conceivable, defeats the knowledge argument" (Streitfeld, Philpapers post 3565).

If you accept the knowledge argument, you cannot support the second premise of the conceivability argument, and vice versa. Therefore, the knowledge and conceivability arguments cannot both be sound. At least one has to go.

References

Tuesday, April 20, 2010

Why the Ability Hypothesis is Important

The ability hypothesis is a defense of physicalism, and it is most commonly discussed as a response to the knowledge argument. Physicalism is the view that the world is entirely physical--or, to put it another way, that all of the physical facts about the world are all of the facts about the world. Against physicalism, the knowledge argument argues that one can know all of the physical facts without knowing all of the facts. This is most popularly discussed in terms of Mary, a super-scientist who comes to learn all of the physical facts about color vision without ever seeing a color. She lives her life confined to a black-and-white room. Yet, intuition tells us that, even though she knows all of the physical facts, she does not know what it is like to see colors. When she leaves her black-and-white room and sees colors for the first time, she learns what it is like to see colors. Thus, she learns new facts--facts which obviously cannot be physical facts, because Mary had already learned all of the physical facts from inside her black-and-white room. Thus, physicalism is false. All of the physical facts are not all of the facts.

According to the ability hypothesis, Mary does not gain new factual knowledge, but only a new understanding of color vision which is constituted by certain abilities--namely, the abilities to remember, recognize, and imagine colors. These new abilities constitute phenomenal knowledge, which is what we mean when we talk of "what it is like to x." The point is that this sort of knowledge is not factual; rather, it is non-propositional know-how.

The knowledge argument and the ability hypothesis are both actively debated among professional philosophers. For example, Sam Coleman recently published a paper arguing that the ability hypothesis is irrelevant and is best forgotten. However, I think I did a pretty good job of disputing Mr. Coleman at PhilPapers.

Another discussion of the knowledge argument has begun at PhilPapers, and I have been defending the ability hypothesis yet again. I recently submitted the following content (with minor adjustments), in which I summarize my views and make a case for the ability hypothesis. My main goal here is to explain why I think the ability hypothesis is important. But first I want to get a clearer picture of the knowledge argument.

First, let's consider Dennett's claim that the knowledge argument is just an "intuition pump," and not a plausible argument. His main point of contention is with the premise that Mary has all of the physical facts. Dennett does not reject this premise; he just says that we cannot use it to draw any conclusions, because we have no idea what it means. It's hard to argue with Dennett here, though many have certainly tried. Ultimately, as Dennett says, it just comes down to competing intuitions.

In trying to understand these competing intuitions, I find it helpful to distinguish between abstract and concrete physical facts. This distinction can probably be construed in a number of ways, but for our purposes, I think it can be put this way: Abstract physical facts are theoretical facts which define relationships between scientifically discoverable entities, whereas concrete physical facts describe particular entities, events or processes. Abstract facts are used to make predictions about concrete events. This is why we say science is a fundamentally predictive activity.

Now, having all of the abstract facts means one can make every possible correct prediction about the concrete physical facts. (If we allow for quantum indeterminacy, then having all of the abstract facts does not equate to predictive infallibility.) A person can have all of the abstract physical facts without having all of the concrete physical facts. The knowledge argument stipulates that Mary learns a completed physics (or a completed science, if we do not want to be reductionists about, for example, biology). It follows that Mary has theoretically unlimited predictive powers. However, it does not follow that she has practically unlimited predictive powers. So there may of necessity be physical facts which she cannot ever know. A physicalist has no problem with Mary becoming master of a completed science and still having an indefinite number of physical facts to learn. Thus, the fact that she learns new facts when she leaves her black-and-white room is not a problem for physicalism.

Furthermore, we need not suppose that there is such a thing as "all of the concrete physical facts." Indeed, it is reasonable to suppose that nobody could ever have all of the concrete facts, because that would require omniscience and the inability to learn from new experiences. If Mary had all of the concrete facts, she could never be surprised by any concrete event. She could not think about what to do, because she would already know what she was going to do. Such a person is inconceivable. A being who could not learn from an experience would not be a person under any common sense of the term. So we should promptly reject the premise that Mary can have all of the concrete facts.

Finally, a physicalist need not assume that it is possible to have all of the abstract physical facts. Physicalism does not require the possibility of a completed science, though I grant that many physicalists are open to it.

In sum, we can reject the knowledge argument on the grounds that it is formulated against an impoverished version of physicalism. It is, in effect, a straw man argument.

This is not the end of the knowledge argument, however. We just need to more clearly get at its motivation. To better indicate the driving intuition behind the knowledge argument, I propose the following reformulation: Mary does not learn every physical fact while inside her black-and-white room. Rather, she learns any arbitrarily large set of physical facts about color vision. She learns some vast, though incomplete, set of abstract and concrete facts about color vision. Yet, our intuition still tells us that Mary will not learn what it is like to see colors until she leaves her room. No matter what she learns while insider her room, she will not have a certain sort of phenomenal knowledge. Thus, phenomenal knowledge cannot be a physical fact.

Dennett's objection does not hold against this revised knowledge argument. Yet, our other responses are still on the table. On the one hand, we might still suppose that Mary learns physical facts which can only be learned by direct acquaintance with color vision. Or, we might suppose that the knowledge gained by Mary's color experiences is not propositional knowledge, but non-propositional know-how.

I wonder what sense there is in claiming that there are facts which can only be known through direct acquaintance. The notion of "fact" seems to entail discursive learnability. At least, some philosophers have supposed as much. See, for example, Ryle (1949, Chapter 9). However, perhaps we can adopt a notion of "fact" which does not entail discursive learnability. This is probably an open question, though we should like some explanation for why some factual knowledge is not discursively learnable. More generally, the question is this: Why should any knowledge only be learnable through direct acquaintance, and not description?

Remember that Lewis (1990) shows that the knowledge argument can be formulated against dualism just as easily as it can be formulated against physicalism. Indeed, we can reformulate the knowledge argument this way: Mary does not simply learn physical facts while insider her black-and-white room. She learns every discursively learnable fact--or, let's say she learns any arbitrarily large number of discursively learnable facts. We need not specify that they are physical facts. Yet, our intuition still tells us that, no matter what facts she learns, she will only learn what it is like to see colors when she leaves that room. Thus, phenomenal knowledge is not discursively learnable, be it factual or non-factual, physical or non-physical. If we say that Mary cannot learn the right phenomenal knowledge while insider her black-and-white room, we must ask why that knowledge is not discursively learnable. It is not enough to say that it is not physical, because that does not explain anything. Even a dualist (or an anti-physicalist) needs some answer to the knowledge argument. This is where the value of the ability hypothesis is found.

Consider the relationship between knowledge and experience. As we gain new experiences, we are directly acquainted with new phenomena. Direct acquaintance gives us phenomenal knowledge. According to the ability hypothesis, this knowledge consists in capacities which underlie our descriptive understanding of the world. In other words, the reason why phenomenal knowledge is not discursively learnable is because it is a set of capacities which are more primitive than our descriptive understanding of the world and experience. I think Ryle (1949) and Wittgenstein (1953) showed rather well that factual knowledge relies upon a sort of knowing which is not factual, but which is manifest in our abilities to follow rules and procedures. The fact that we can follow rules and procedures for identifying colors means that our understanding of color vision is not wholly propositional, or factual. To put it another way, the fact that we can follow rules about how to identify colors is evidence that we have an understanding of color vision which does not itself depend upon factual knowledge--and which itself cannot be discursively learned. (As I noted in a post in another thread, Ruth Millikan may also be counted as a supporter of this view, even if she never to my knowledge explicitly remarked on the ability hypothesis.)

In conclusion, the ability hypothesis is called upon to explain why phenomenal knowledge is not discursively learnable. This explanation is required for dualists (or, more generally, non-physicalists) as well as for physicalists. Furthermore, the ability hypothesis accounts for the intuition motivating the knowledge argument without abandoning physicalism. The ability hypothesis is a physicalist doctrine, because it defines phenomenal knowledge wholly in relation to physical processes, events and capacities. Therefore, in accounting for the intuition motivating the knowledge argument, the ability hypothesis offers explanatory power in favor of physicalism.

References

Gilbert Ryle (1949/2002). The Concept of Mind. Hutchinson and Co.

David Lewis (1990). "What Experience Teaches." In William G. Lycan (ed.), Mind and Cognition. Blackwell.

Ludwig Wittgenstein (1953/1968). Philosophical Investigations. New York,Macmillan.