1.0
Douglas Alan
24.500: Consciousness
Prof. Alex Byrne
18 May 1998
2.0
Chalmers' first argument against physicalism attempts to persuade us
that there is a possible world that is a minimal physical duplicate of
our world, yet is phenomenally void---that is, it is a zombie world.
This argument relies on Chalmers' ``conceivability/possibility
thesis'', or ``CPT'' for short. CPT is the thesis that all ``ideally
conceivable'' (or ``conceivable-in-principle'') worlds are possible
worlds.
If CPT
is true, and it is true that we can conceive of a phenomenally void
minimal physical duplicate of our world, then there is such a possible
world, and consequently physicalism is false. Hill and McLaughlin
argue that CPT is false, and that it is an a posteriori fact that any
minimal physical duplicate of our world is exactly like ours
phenomenally.
For the purposes of this paper I am going to accept much of Chalmers' 2D framework. That is, I accept that a term of English and its Twenglish counterpart share an aspect of meaning (or ``primary intensions'', to use Chalmers' terminology), even though the English word ``water'' and the Twenglish word ``water'' refer to different things under all circumstances (or have different ``secondary intensions''). I will bring up differences between my view and Chalmers' where they occur in the dialectic. I hope to show that one can be quite sympathetic to Chalmers' approach, yet not end up with his conclusions.
A different approach to counter Chalmers is to reject his 2D framework outright. You might think, for instance, that a term of English shares no aspect of meaning with its Twenglish counterpart, and therefore that talk of primary vs. secondary intensions, or of strong necessities being different from Kripkean necessities, is nonsensical. If you think this way, you might not object too loudly, however, if it can be shown that Chalmers can be defeated in his own back yard, so to speak.
I also think that much of the debate might be carried on with fewer contentious notions. One might allow that a term of English shares a ``sense'' with its Twenglish counterpart, even if it shares no aspect of meaning. And talk of concepts might be cashed out in terms of a collection of abilities possessed by a speaker who uses a term. I won't attempt to carry out this project here.
Conceivability clarified?
Just what is meant by ``conceivable'', much less ``ideally conceivable'', has been left a bit unclear by Chalmers. Hill and McLaughlin (borrowing in part from Thomas Nagel) try to clarify the issue by identifying four different types of conceiving. First they split conceiving into ``imagining'' and into ``conceiving of''. To imagine is to construct within oneself a qualitative or imagistic mental representation. To conceive of is a bit more abstract: it is to construct within oneself a mental ``representation that has a logical structure and that has concepts as its basic building blocks''. This clarification has not been contested by Chalmers.
Two different kinds of imagining are then identified: ``imagining perceptually'' and ``imagining sympathetically''. To imagine something perceptually is to construct a mental image of the imagined thing. For instance, to imagine an apple, you would make a mental image of an apple. To imagine something sympathetically is to put yourself in state resembling the thing itself. For instance, to imagine Otto's anger, you might think of Otto while also making yourself a bit angry. Two different kinds of conceiving of are identified: conceiving of using phenomenal concepts and conceiving of using physical concepts. Hill and McLaughlin also introduce the term ``Cartesian intuitions'' to abbreviate ``modal intuitions that seem to support property dualism'', e.g., intuitions derived from imagining zombie worlds.
The first thing that comes to mind is that I am not sure this really makes anything clearer. Just what concepts are is murky enough, but phenomenal concepts are even murkier. Hill and McLaughlin claim that they are recognitional concepts, but this is sure to be highly contentious. Furthermore, it is not obvious just what ``a mental representation that has a logical structure with concepts as its basic building blocks'' is supposed to be. A sentence, for instance, seems to be a logical structure with terms standing in for concepts, and therefore, if I hold a sentence in my head, according to the explication provided by Hill and McLaughlin, have I conceived of world in which the sentence is true?
If this view is allowed, it seems that I can conceive of a world with a round square, since I can certainly hold the sentence ``There is a round square'' in my head. But if CPT is to be taken at all seriously, we have to say that I can't conceive of a world with a round square. We might want to say that I can tentatively conceive of a round square, but this conception is quickly blocked when I notice the contradiction. If we take this position, however, it seems that for any proposition, no matter how nonsensical, I can conceive of a world where it is true, as long as the proposition can be expressed as an English sentence. The conception will then be blocked only when a contradiction is noticed.
If this is the case, however, saying that we can conceive of a world is no different from saying that the world can't be ruled out a priori. And if this is the case, then Chalmers' argument from conceivability just boils down to the same thing as the knowledge argument. We might as well, then, just dispense with the conceivability argument as being redundant and more confusing. Consequently, the exposition by Hill and McLaughlin on conceiving of seems to leave us with a notion of conceivability that is either pointless for this debate or provides no more enlightenment than that with which we started. I'll assume the latter, and press on, working on the questionable assumption that even if we don't understand what we are doing when we are conceiving of, we know when are doing it.
Imagining seems less problematic than conceiving of, but it too has its problems. Imagining is less problematic because mental images are much more constrained than sentences are in what they might propose. While it seems that I can at least tentatively conceive of a round square, I don't think I can tentatively imagine one. What I can visually imagine seems to roughly correspond to what I might be able to draw or paint if I were to have any artistic skills, and I know of no one who can draw a round square. There are some things, however, that I can at least tentatively imagine that are not possible: the Penrose staircase and Penrose tribar made famous by M.C. Escher come immediately to mind. Is it supposed to be the case that noticing a contradiction in these things makes them unimaginable?
Another problem is that according to the exposition by Hill, McLaughlin, and Nagel, it seems that I should be able to imagine---at least tentatively---an angry square. I should be able to do this by visualizing a square and then making myself a bit angry. Alas, I'm not able to get even to the tentative imagination stage. The best I seem to be able to do is to imagine that I'm angry while looking at a square.
Despite the fact that Hill and McLaughlin's putative clarification seems quite problematic, let's accept it for the time being and see where it gets us. We shall see that it doesn't really get us very far.
Just how much can we trust our Cartesian intuitions?
Hill and McLaughlin argue that although conceivability involving a
single type of conceiving might be a perfectly reliable guide to
possibility, conceivability involving several different types of
conceiving might be wildly inaccurate. Specifically, imagining
a zombie world involves both perceptual imagination and sympathetic
imagination. Conceiving of a zombie world involves both
phenomenal concepts and physical concepts. Our modal intuitions in these
cases, they say, cannot be trusted because cognitively these
intuitions are very different from other kinds of modal intuitions
that Hill and McLaughlin will grant to be trustworthy.
My response is to agree that the question of when and to what degree we should trust our intuitions is very vexed, more so than Chalmers seems to acknowledge. If we could practice philosophy using only forms of reason known to be perfectly reliable, then it is clear that that is what we should do. Unfortunately, much of human knowledge seems to be only available implicitly---via intuitions and capabilities that we do not have the ability to formalize. Although it may be worrisome to rely on reasoning that cannot be fully expressed and is sometimes incorrect, I think that without relying on our intuitions, philosophy would make no progress at all.
Furthermore, I think that our intuitions are of a myriad kind. If
each intuition had to be individually judged as to whether or not it
were cognitively similar enough to other sorts of intuitions that we
already have reason to believe are reliable, we might never accept any
intuitions at all. Hill and McLaughlin's
argument might make a very interesting postmortem if we were to
conclude for other reasons that we should reject Cartesian intuitions,
but I think that the argument is not strong enough by far to be a good
reason to reject Cartesian intuitions outright. As with other
intuitions, it still seems quite reasonable to assume---perhaps with a
pinch of salt---that what our Cartesian intuitions tell us is correct,
until given good reason to believe otherwise.
Chalmers' argument for CPT
Hill and McLaughlin's case against Cartesian intuitions does not rest there. They take up the more general issue of CPT. Chalmers, they say, does not argue for CPT; rather, he just seems to assume that it is true. The only defense he makes of CPT is to explain away apparent counterexamples. Explaining away counterexamples, they say, does not count as a defense of CPT. A defense of CPT ``would have to show that the principle systematizes our modal intuitions, and it would also have to show that the principle satisfies a general coherence requirement.''
Chalmers' reply is that there are a number of positive reasons for
accepting CPT: (1) Without it there will be ``strong
necessities'', which
will be brute and inexplicable. (2) If CPT is rejected, it will lead
to an ad hoc proliferation of modalities. I.e., philosophers will
need to split the space of worlds into at least two spaces -- the
space of metaphysically possible worlds (those worlds that really are
possible) and a larger space of ``logically possible worlds'', which
would include all metaphysically possible worlds, plus the worlds that
are conceivable but not truly possible. (3) The acceptance of strong
necessities ``would lead to a philosophical revolution far more
radical than Kripke's''.
Chalmers does have a point in that his model seems simpler and more elegant than the alternative. I don't know, however, if it lives up to the demands of Hill and McLaughlin for ``systemization'' and satisfaction of a ``coherence requirement''. A friend of Chalmers might point out that in the field of science simplicity and elegance alone are considered virtues enough to accept a theory, as long as the theory stands up to all potential counterexamples. Why shouldn't this scientific principle work just as well in philosophy?
Unfortunately, I don't think that this principle would work nearly as well in philosophy. Science doesn't need to be as wary of strong claims because in the realm of science strong claims are usually disproved with relative ease if they are grossly out of accordance with reality. In the realm of philosophy, however, matters are quite different---how to interpret putative counterexamples is not so clear, while in physics, for instance, it is relatively straight forward. In philosophy, the debate of how we should interpret a putative counterexample might rage for a thousand years, while in physics a counterexample will make it clear that the theory is wrong and needs to be repaired or scrapped. A theory in philosophy can seem wonderfully beautiful, but make such strong and unverifiable claims that we are very loath to accept the theory without having some more certain way of knowing that it is true.
An example of such a theory is David Lewis's thesis of modal realism,
the thesis that all possible worlds exist and are the same kind of
thing as the actual world. That is, a possible world is a large
physical object, and the
inhabitants thereof are just as real as you and I. The actual world
enjoys no special status compared to other possible worlds---the
actual world is just the possible world we happen to find ourselves
in. Modal realism offers a more simple and elegant view than we might
otherwise have, and it also solves some eternally vexing philosophical
problems. If we accept modal realism, for instance, not only does the
mystery of what possibility really is disappear, but we also
come to know why there is something rather than nothing. On the other
hand, the primary reason that most of us don't accept modal realism
is, I presume, that the ontological cost is just too staggering: Do we
really want to believe that the number of sentient beings there are is
so much larger than we ever previously imagined? Do we want to
believe this just so we can make our theories a bit more
elegant?
I should also point out that scientists are not always so eager to accept theories only on the basis of simplicity and elegance---particularly when the theory is very difficult or impossible to verify and the ontological cost is high. Talk of virtual particles was for a long time, I believe, often felt to be talk of nonexistent fictions that helped make the math easier. And the multiple-worlds interpretation of quantum mechanics has not been widely accepted, even though it is a more elegant theory than the alternatives, due to, I assume, the high ontological cost.
The ontological cost for accepting Chalmers' CPT is not as large as
for accepting modal realism, but it is significant. Unfortunately, I
find disputes of this sort quite difficult to adjudicate, so I'll just
say that in my opinion the cost is large enough that we should refrain
from accepting CPT on the basis of simplicity alone. On the other
hand, simplicity and elegance are features that we desire in our
theories, so I do think CPT deserves serious consideration. As things
stand in the dialectic, we should leave it on the table pending more
evidence either way.
Epiphenomenalism and panprotopsychism
Hill and McLaughlin continue with their case against CPT by asserting that instead of making things simpler and more elegant, it would force us to conclude that the fundamental laws of our world are not elegant as we had previously supposed; rather, they would have to include an ``independent swarming mass of fundamental psychophysical laws''. This is a surprising claim for Hill and McLaughlin to make without any argumentation, since Chalmers in The Conscious Mind clearly addresses this objection. The strategy that Chalmers says he is most drawn to is panprotopsychism, the view that fundamental particles have intrinsic properties (as opposed to only the extrinsic properties specified by physics) and that some or all of these properties are protophenomenal. According to this view, there are some new fundamental laws, but such laws would not necessarily be numerous or complex. Chalmers raises the potential objection that with panprotopsychism protophenomenal properties seem to ```dangle' ontologically from physical laws'', but he handles the objection by arguing that that the basic properties of the world might be neither physical nor protophenomenal; rather, both might arise out of some even more basic properties. Some philosophers may claim that panprotopsychism gives them the creeps, but I don't have much to say to this---I don't really understand this feeling.
Hill and McLaughlin assert that CPT implies epiphenomenalism, the
thesis that one's phenomenal states exert no causal
influence. Epiphenomenalism, they claim is highly
counterintuitive. And of course it is. But I don't see that it is
any more counterintuitive than the status of free will in a universe we
believe to be deterministic. The issues involving
epiphenomenalism and free will seem very similar to me, and if we
think we have good reason to believe that our intuitions about free
will are very wrong, then it seems that very similar reasons might
lead us to believe that our intuitions about epiphenomenalism are
wrong. Also, once again, Hill and McLaughlin have ignored Chalmers'
way out: panprotopsychism.
Panprotopsychism allows a place for the causal influence of phenomenal
states. As Chalmers points out, however, the amount of causal
influence allowed for with panprotopsychism may be less than we wanted
intuitively. That is, a zombie duplicate of you would behave the same
way as you under all circumstances. Our intuitions, in contrast, tell
us that our phenomenal states are an essential component in the causal
network that results in our cognition and behavior. That is, our
intuitions tell us that duplication of our phenomenal states would be
required to duplicate our cognition and behavior.
It should be noted, however, that psychophysical identification, which
is proposed by Hill and McLaughlin as the best alternative to
Chalmers' position, suffers the same problem with respect to
phenomenal causation as panprotopsychism: it allows for a behavioral
and cognitive duplicate of you that is phenomenally void. To see
this, consider that if physicalism is true, it is possible to
construct a functional duplicate of you using a digital computer that
would duplicate your cognition and behavior. Now, let's say that Hill and McLaughlin identify
pain with C-fiber stimulation. According to this theory, your
functional duplicate would not feel any pain because it has no
C-fibers, only simulated C-fibers. Consequently, psychophysical
identification, like panprotopsychism, provides a weaker kind of
causation for phenomenal states than we intuitively would like. Only
identification of phenomenal states with functional states can restore
phenomenal states to being an essential component in the explanation
of cognition and behavior.
The counterintuitive consequences of psychophysical identification
Psychophysical identification has a number of consequences that are quite counterintuitive, despite Hill and McLaughlin's assertions to the contrary. Since their argument rests, in part, on the proposition that psychophysical identification has consequences that are far less counterintuitive than those of Chalmers' view, the counterintuitive consequences of psychophysical identification need to be carefully weighed.
One such counterintuitive consequence is, as Jaegwon Kim points out,
that it denies that octopi suffer pain if they don't happen to have
C-fibers. And if we are going to deny any commonality between human
pain and whatever it is octopi feel when they appear to be in pain, it
is not clear why we should grant a commonality between what I feel,
what you feel, and what Otto feels when we each say we are in pain.
If we are going to so severely differentiate human pain from octopi
phenomenal states, why not also make a severe differentiation between
the phenomenal states of individual humans? That is, why not say that
I feel pain
, you feel pain
, and Otto feels pain
, and there
is nothing in common between these things? Or consider that we might
slowly replace the neurons in your C-fibers with functional duplicates
made out of silicon. Your pain would slowly fade away without you
ever noticing---all the while you would claim that your pain hurts
just as much as always, but eventually you would really feel none. In
contrast, your ability to feel pleasure, or any other sensation, would
remain untouched.
Another counterintuitive aspect of psychophysical identification is
that it appears that a proponent must either accept a type of
functionalism or accept something like panprotopsychism. To see this,
consider the following possible worlds: (1) A world where the laws of
physics are exactly the same as in ours and containing people just
like us. (2) A world where the laws of physics are exactly the same
as in ours and containing people just like us, except that in this
world all the particles have been replaced by their antiparticles.
(3) A world in which the laws of physics appear to be just like
ours, except that in this world electrons and quarks are not
fundamental particles. Instead, an electron is made out of
thirty-three fundamental particles and a quark is made out of
forty-two fundamental particles. The inhabitants of this world,
however, will never know that this is the case, because the energy
required to split the electron or quark is greater than the total
amount of energy in the entire world. (4) A world in which the first
thing God did was to set the physical law to be completely unlike that
which is true in our world. He then constructed in this world a huge
digital computer with incredible computing power and programmed it to
``perfectly simulate'' the physical laws of our world. Finally, he fed into the computer the
initial conditions (e.g., the position and momentum of every
fundamental particle in our world at the beginning of time) and set
the computer running.
The first thing that one should notice about these worlds is that they
are all epistemically identical. That is, there is no way for
an inhabitant of one of these worlds to tell which of these worlds
they are in. There is simply no experiment that can be performed to
distinguish between the possibilities. In fact, all the worlds, if they
start off with matching initial conditions (and we ignore complications
due to quantum randomness), will march down time together, forever in
perfect synchrony. Or to think about this a slightly different way,
consider the possibility that we might someday open up a
communications channel to a parallel universe that is just like one of
these worlds. Through this communications channel we fax the
inhabitants copies of our finest physics text books and they fax us
copies of theirs. Neither we nor they will notice any discrepancies
between our theories of physics. We'd like to know whether the
inhabitants of the parallel universe are made out of matter or
antimatter, or whether the things they call electrons are really a
swarm of 33 particles, or whether their electrons might really be
computational states in some sort of cosmic computer. There is no way
for us to answer these questions. There is simply no experiment that
we can devise for them to perform that will confirm or deny the thesis
that what we call electrons are the same things that they call
electrons.
So what is a psychophysical identificationist to say about this? I can think of only two plausible approaches: The first approach is to say that the differences between these possible worlds are not relevant to any higher level facts. That is, physicalism presupposes that it doesn't matter what fills the roles of electrons and quarks, as long as the objects in question behave the way physics says they should (i.e., that they properly fill the electron-role and quark-role that have been specified functionally). It seems to me, however, that whatever the reasons were to reject functionalism about phenomenal states at a higher-level, these reasons would be just as convincing against functionalism at this lower level. Consequently, the psychophysical identificationist will be forced take the second approach: He will conclude that there are no pains in worlds (2), (3), and (4). Our duplicates living in these worlds will have no pains because they must possess C-fibers in order to feel pain. C-fibers are made out of electrons and quarks. The electron and quark analogues in the alternate worlds do not qualify as electrons and quarks, and consequently there are no C-fibers in these worlds.
This conclusion leaves the psychophysical identificationist in a dilemma, however, since if this is the case, physics, as we will ever be able to understand it, does not fix the phenomenal facts. The inhabitants of the other worlds will have the very same theory of physics, yet even if our universe and their universe appear to be in the very same physical state, phenomenal facts will differ between our universe and theirs. This would be a hollow victory for physicalism (it is very similar to Chalmers' panprotopsychism), and I gather that the the psychophysical identificationist won't want to accept this view either. If he were to accept this view, Chalmers and the psychophysical identificationist would be largely in agreement on this particular issue and would just be arguing at cross-purposes.
Due to the above considerations, I think that there is a very strong
case that contra Hill and McLaughlin, if we are to reject
panprotopsychism, the simplest and most coherent alternative
explanation for the correlation of phenomenal states with their
nomologically correlated brain states is functional identification,
not psychophysical identification. I take it that functional identification is
often rejected due to the conceivability of functional duplicates that
are phenomenally void, but if Cartesian intuitions have been rejected
for physical duplicates, it is hard to see why we should then accept
Cartesian intuitions for functional duplicates.
Recognitional concepts
Hill and McLaughlin's final argument is, they say, a bit of a detour and not directly related to CPT. We shall see that their last argument is actually quite apropos. It addresses a footnote by Chalmers in The Conscious Mind:
Loar suggests that phenomenal concepts are recognitional concepts, and argues that it is reasonable to expect a recognitional concept R to ``introduce'' the same property as a theoretically specified property P. He gives the example of someone who is able to recognize certain cacti in the California desert without having theoretical knowledge about them. But this seems wrong: if the subject cannot know that R is P a priori, then reference to R and P is fixed in different ways and the reference-fixing intensions can come apart in certain conceivable situations. Unless we invoke the additional machinery of strong metaphysical necessity, the difference in primary intensions will correspond to a difference in reference-fixing properties. [p. 373]
Hill and McLaughlin conclude that the following premise is being relied on by Chalmers:
I don't think that this is quite what Chalmers had in mind, though I see how Hill and McLaughlin could have come to this conclusion---the paragraph by Chalmers is somewhat ambiguous. I think Chalmers was using the following premise instead:
)
Chalmers claims that (
) is a consequence of CPT.
Presumably, if (
) is false, the subject can conceive of an
X being not-Y even if X and Y have the same
primary intension and hence the same extension in all possible worlds.
This is clearly not allowed by CPT. But if this is what Chalmers has
in mind, then the conceivability of a world seems to amount to nothing
more than the claim that it can't be ruled out a priori. And if this
is the case, as I pointed out earlier, we should just discard the
argument from conceivability in favor of the knowledge argument.
On the other hand, most of the time Chalmers seems to have in mind a
stronger notion when he talks of conceivability. In this case, (
)
might fail without taking CPT with it, but (
) certainly is a
consequence of Chalmers' premise that there are no strong necessities.
If (
) fails, it thereby takes the foundation of CPT with it; that
is, if we come to decide that there are strong necessities, Chalmers'
simple and elegant view has been destroyed, and consequently we no
longer have a reason to accept CPT.
Hill and McLaughlin's response is to (a), not (
), but I
don't think this matters since it is just as relevant to (
)
as it is to (a). Their response is to argue that it is plausible
that the reference of our concept of pain is fixed by the fact that we
can recognize when we are in a certain mental state, and the reference
of our concept of C-fiber stimulation is fixed by a certain a
posteriori theory of neuropsychology. These concepts, say Hill and
McLaughlin, refer to the very same state, but we can't tell a priori
that they do.
Although Hill and McLaughlin don't mention it, their argument does not depend on psychophysical identification. A functionalist will object to the identification of pain with C-fiber in Hill and McLaughlin's argument, but she can otherwise accept it. Instead of identifying pain with C-fiber stimulation, the functionalist will identify pain with a certain functional state (say ``state-P''), but the point will remain the same: A recognitional concept (that of pain, for instance) can have the same primary intension as a theoretical concept (state-P), yet the fact that the primary intensions are identical will not be knowable a priori.
Chalmers' reply is to say that this, once again, comes down to whether or not there are strong necessities---Hill and McLaughlin haven't provided a good enough reason to overturn his beautiful theory. And this is true---Hill and McLaughlin don't do much to argue that pain is a recognitional concept, and they don't do much to argue why we should believe that it isn't a priori that pain and C-fiber stimulation have the same primary intension; they only argue that their position is plausible.
I think a better case can be made. There are very good reasons to believe that some of our concepts are recognitional concepts, and that the existence of recognitional concepts implies the existence of strong necessities. I will elaborate with an example, but because phenomenal concepts are quite difficult to think about, and because their nature is quite contentious, and because it's not even clear what they are, the example will involve only a nonphenomenal concept. It is not phenomenal concepts per se that imply strong necessities---any recognitional concept will do.
Consider my ability to recognize syntactically correct sentences. I have no idea how I do this---I just look at or hear a sentence and somehow I know whether or not it is syntactically correct. Now suppose that I become the subject of a scientific inquiry into my linguistic skills. With the help of an MRI machine and some radioactive tracers, a team of cognitive scientists are able to determine the computational process that my brain uses to determine whether or not a sentence is syntactically correct.
Let us suppose that the cognitive scientists don't tell me just what they are up to. Instead, they hand me a printout of a computer program that implements the computational process without any explanation of where this program came from or what it's supposed to do. They do tell me that the program takes as its input a finite sequence of words, and that its output is either ``true'' or ``false''. They ask me to memorize the program and to mentally execute it on whatever sequences of words I happen to come across in my daily affairs.
Can I tell a priori that the program outputs ``true'' just in case I will find the sentence to be syntactically correct? I think not. After a while I might notice an unwavering correlation, but I can't rightly conclude with certainty from this correlation that I and the program will always agree. The possibility will always remain, as far as I will be able to tell, that someday I might stumble across a sentence where the program and I disagree. Since there is no difference between the primary and secondary intensions of the term ``syntactically correct sentence'' or of the program, which when memorized provides a theoretical concept of a syntactically correct sentence, we have found a strong necessity---the primary intension of my concept of a syntactically correct sentence and the primary intension of the program are the same, yet it is not a priori that this is the case.
Chalmers might reply that I should be able to intuit the details of the computational process that my brain naturally uses to parse sentences. But to assume that this is the case is to assume that I have some special powers of divination into the construction of my own mind. We have no reason, other than Chalmers' idea of modal simplicity, to believe that I, or any of us, should have such powers. I think we can see now that CPT is not really just a theory about abstract modal spaces, as Chalmers innocently claims it to be---it is really a theory about the human mind. And it is not a theory that would simplify our view of the human mind; rather, it would make it more complicated. We would now have to posit cognitive structures in the human brain that exist just to make philosophy easier, and with no other apparent purpose.
Chalmers might object that the reason I can't figure out a priori that my concept of syntactical correctness and the program will always pick out the same sentences is because I do not have the superhuman abilities required for ``ideal conceivability''. Grant me these superhuman abilities and I could consider every possible finite sequence of words (all infinity of them) and compare what my recognitional concept tells me about each sequence with what the program tells me. If the two different concepts give me the same answer for every element of this infinite set, then I would have determined a priori that the two concepts share the same primary intension. But, as I shall argue more thoroughly in a few paragraphs, if ideal conceivability requires such superhuman capabilities as this, it is not clear of what use this notion of ideal conceivability is supposed to be to us mere mortals.
Another possible reply from Chalmers is that linguistics is an a priori venture, so if I were smart enough, I would be able to come up with a good linguistic theory, and then I'd be able to notice that the program provided by the cognitive scientists is in accord with my linguistic theory. Linguists, however, despite many years of trying have yet to come up with a grammar for English that works well. Psycholinguists believe that understanding how we parse sentences will require a great deal more a posteriori research into the workings of the human mind than linguists previously supposed, and it was never a completely a priori venture to begin with.
Also there are many other examples of recognitional concepts. The
idea of syntactical correctness is far from the only one. For
instance, Otto might be very good at recognizing a species of rare
cactus, without knowing how he does it. And most of us are very good
at recognizing faces and recognizing emotions on faces, but we don't
know he we do these things. If we did, software designed by AI
researchers to perform these tasks would, no doubt, work much better
than it does.
Each recognitional concept you possess allows you to identify a class C of objects recognitionally on the basis of an ability implemented by your brain. All the scientific evidence indicates that the intension determined by this ability will not always be available a priori. So, if a recognitional ability of yours is implemented computationally and consequently something is a member of C if and only if a given computer program P says it is, P will not always be knowable a priori. In this case, you won't be able to come up with P a priori, nor will you be able, given P, to tell a priori that P picks out C.
If the construction of the human mind were knowable a priori,
the entire field of cognitive psychology would appear to be an
unnecessary doctrine. Philosophers would be able to determine how the
mind works without any empirical research program. And if the construction
of the human mind is not knowable a priori, then we should only expect
that that there will be concepts for which the primary intension is
not knowable a priori. (Though I suppose we can't completely rule out
for this reason alone the unlikely possibility that just by
coincidence all concepts have a priori knowable primary intensions,
while the a posteriori nature of the human mind is completely
contained to other aspects of cognition, it seems like we'd need a
very good reason to believe that this remote possibility turns out to
be the case.)
To sum up, I think it is clear that recognitional concepts exist, and this is enough to establish the existence of strong necessities. The existence of strong necessities leaves us without a reason to believe in CPT. So, although Hill and McLaughlin got off to a slow start, their final argument, when elaborated, leaves CPT on very shaky ground. But we don't have to stop there---some additional reasons to doubt CPT come to mind.
Misdescribing possible worlds
Chalmers lets slip a huge gap in CPT when he addresses the issue of Goldbach's conjecture. (Goldbach's conjecture is the claim that every even natural number is equal to the sum of two primes.) There are no known counterexamples to Goldbach's conjecture, but it has not been proven true either. It seems that we can conceive of a possible world where Goldbach's conjecture is proven false, but it also seems that we can conceive of a world where it is proven true. At least one of these conceivable worlds must not really be possible. (Both worlds might be impossible if it turns out to be true but unprovable.)
Chalmers addresses this apparent problem for CPT by claiming that both of these worlds are possible worlds, but one of them has been misdescribed: in one of the worlds, a mathematician has made a mistake and published an invalid proof. Allowing this, however, is a can of worms. Prior to this example, Chalmers had an elegant story to tell about the ways in which conceivability could go wrong---there was only one way: any misdescription was due to Kripkean necessities.
For instance, you might conceive of a world in which water is not
H
O. But it is a Kripkean necessity that water is H
O, so such
a world is not possible. CPT claims that you did conceive of a
possible world, but it is a world in which watery
stuff is not
H
O, and you misdescribed it as a world where water is not
H
O. Misdescriptions due to Kripkean necessities turned out to be
unimportant for the philosophical issues at hand.
But now Chalmers is allowing other sorts of misdescriptions into our conceived worlds and it is unclear just what the ramifications are. Perhaps when we conceive of a physical duplicate of our world that is a zombie world, we have misdescribed the conceived world in an important way. Perhaps it is not really a physical duplicate of our world after all. Or perhaps it is not really a zombie world. That is, perhaps the world has been misdescribed due to a modal fallacy: it is certainly possible that a world might be a physical duplicate of our world, and it is certainly possible that a world might be phenomenally void. A world that is attributed both properties, however, might be misdescribed.
Thinking about Goldbach's conjecture has given us some doubt about CPT. No longer do all misdescriptions of conceived worlds occur due to a single well-understood cause (i.e., Kripkean necessities). Unless an account of precisely when else misdescriptions may occur, there will always be the lingering suspicion that perhaps our imagined zombie world has been misdescribed in some manner. Chalmers might offer up such an account: Because we are able to conceive of a world in which Goldbach's conjecture is true, and we are also able to conceive of a world in which it is false, it becomes immediately apparent that one of them has been misdescribed. In general, whenever we can conceive of two worlds that differ on a necessary truth, one of them is misdescribed. This issue doesn't come up for the imagined zombie world.
Or does it? Consider now whether we can conceive of a world that is a minimal physical and phenomenal duplicate of the actual world, and in which the phenomenal facts are necessitated by the physical facts. If there is such a possible world, then physicalism is true in the actual world. Now it seems that many philosophers can indeed conceive of such a world. In fact, they conceive of the actual world as being such a world. If they didn't, the thesis of physicalism would probably not be so popular. According to CPT, since this world is conceivable, it is possible, and consequently physicalism is true. Of course, it is immediately apparent there is a contradiction here: the physicalistic world and the zombie world disagree on a necessary truth, so one of them must be misdescribed. Which one? We seem to have no good reason at the moment to prefer one over the the other.
Chalmers will probably reply that this whole issue related to Goldbach's conjecture is a red herring. The problem we have been discussing is a problem with prima facie conceivability, but not with ideal conceivability. Under ideal conceivability, we would know, for instance, whether or not Goldbach's conjecture is true, and thus we wouldn't make such a misdescription of a world. Or perhaps under ideal conceivability we would refrain altogether from conceiving of worlds about which we are missing some important fact that might prevent us from making such a misdescription.
The problem with this reply is that if prima facie conceivability is not a perfect guide to ideal conceivability, and the only thing we are capable of is prima facie conceivability, then it is not clear how conceivability is supposed to be a better guide to answering our philosophical questions than any of our other intuitions. It appears that conceivability intuitions are useful for supporting or attacking beliefs on which more reliable forms of reasoning have so far remained silent. But the same was already true of all our intuitions (or at least those intuitions that are not of a sort already known to be particularly unreliable), so it is not at all clear what new capabilities CPT is supposed to give us.
Perhaps the idea is that conceivability is a more reliable type of intuition than other types. But just how much more reliable? Your guess is as good as mine. Or perhaps the idea is that entire philosophical community working together is supposed to approach the capabilities of a superbeing for whom ideal rational reflection is second nature. If the philosophical community only approaches the capabilities of a superbeing, then there is still the possibility that mistakes will be made, and the extremely difficult issue of phenomenal consciousness seems like a likely area where the community's powers would fall short. On the other hand, if the argument can be made that the philosophical community's capabilities are equal to those of the superbeing, then perhaps this objection can be vanquished. That's a mighty big ``if''.
CPT appears paradoxical
There's one final problem with CPT that I'd like to mention. CPT appears to be paradoxical. I can imagine a minimal physical duplicate of our world that is phenomenally void and for which physicalism is true. Furthermore, I imagine that something very important to my nature is missing from this zombie world (namely my phenomenal states). If CPT is true, physicalism is false in our world, and the zombie world is possible.
In the zombie world, my zombie duplicate will be imagining a possible world that is a minimal physical duplicate of his world, yet where something important to his nature is missing. But because physicalism is true in his world, nothing that exists in his world will be missing from a minimal physical duplicate of his world. Therefore, the world my zombie duplicate imagined is not a possible world, and consequently CPT is false in the zombie world.
But why should CPT be true in our world, yet false in a zombie duplicate world? Do zombies suffer some sort of cognitive impairment that we don't? How could this be? Our zombie duplicates appear to be cognitively identical to us.
Perhaps the way out of the paradox is to claim that CPT is true in the zombie world and my zombie duplicate did imagine a possible world, but he misdescribed the world he imagined: the world he imagined was really his own world, but he misdescribed it as a different world where something important is missing. If this is the case, though, what reason do I have to suppose that I have not misdescribed the world I have imagined. The natural conclusion would be that I have also imagined my own world and misdescribed it.
Chalmers might assert that my zombie duplicate is cognitively impaired
in some way in which he will never be aware, and that despite being
physical and functional duplicates, my zombie duplicate and I are not
cognitive duplicates. I won't examine this issue any more here, but I
suspect that such a reply will not be very convincing.
Can Cartesian intuitions survive the death of CPT?
If I am right, CPT is left in tatters. But perhaps all is not lost
for the neo-Cartesian. Cartesian intuitions seem to be perfectly good
intuitions that deserve to be taken seriously. The antiphysicalist
will not have CPT to raise these intuitions to the level of certainty,
but we haven't yet seen a knock-down argument that would cause us to
reject our Cartesian intuitions either. Hill and
McLaughlin argued that it is plausible that phenomenal concepts
are recognitional concepts, but this is not enough to establish that
phenomenal concepts are in fact recognitional. Such a claim is
sure to be hotly contested. And even if phenomenal concepts are
recognitional concepts, this does not imply that Cartesian intuitions
are false. For all that's been said, it might still be the case that
the properties recognized by phenomenal concepts are not compatible
with physicalism.
The antiphysicalist might try to increase our confidence in Cartesian intuitions by arguing that they are similar to other modal intuitions that we feel to be quite reliable. Imagine, for instance, says the antiphysicalist, that in the near future we will all come to know that there are ectoplasmic angels (perhaps we will discover an irrefutable message from God telling us so encoded in the cosmic background radiation). Despite this knowledge, some philosophers remain physicalists. They claim that ectoplasmic angels are not a problem for physicalism because facts about ectoplasmic angels supervene on the physical facts---they are necessary but a posteriori consequences of the physical facts.
Clearly, if this scenario comes true, then physicalism will turn out to be false and the aforementioned physicalists are wrong. How do we know they are wrong? Because we can conceive of a physical duplicate of our world without the angels. Perhaps there are times when our ability to conceive fails us and leads us astray, but this is not one of those times. We can tell it is not one of these times because it is just so obvious that ectoplasmic angels are so qualitatively remote from anything that could be a consequence of physical law that we can immediately eliminate the possibility that they are. We can depend on our Cartesian intuitions for the same reason---they are just like our modal intuitions about ectoplasmic angels. Phenomenal states are obviously so qualitatively remote from anything that could be a consequence of physical law that we can immediately eliminate the possibility that they are.
This argument sounds good on first blush, but let's consider again what's happening with the zombie duplicate of the antiphysicalist. His intuitions about ectoplasmic angels are just as accurate as ours, but what about his Cartesian intuitions? Alas, his Cartesian intuitions are completely wrong---for him, physicalism is true despite his Cartesian intuitions. Thus, it appears that Cartesian intuitions must differ in some important way from angelic intuitions, and consequently we end up where we started from, with the nagging suspicion that physicalism is false, but without a firm reason to confirm our suspicions.
