Posted Tue Feb  6 08:30 GMT 1996

Newsgroups: sci.psychology.consciousness
References: <Pine.3.89.9602021302.A14932-0100000@world.std.com>
Subject: Re: Tucson II [Function and Experience]

Much of what I read about consciousness, even when written by eminent
scientists and philosophers, seems to me to involve a naive and
unjustified assumption that just because we are normal human beings we
have some clear notion of "consciousness" that we can use in asking
philosophical questions, or worse, asking what are supposed to be
scientific questions (e.g. "How did consciousness evolve?" or "What's
the function of consciousness?" or "Is consciousness necessary for
thought?").

By contrast, the recent posting by Thomas Clark <twc@world.std.com>
implies that we may have to reject the naive assumption that
we know what we are talking about just because we think we do.

I agree with almost everything he wrote. I'll pick out some points for
comment and elaboration.

> ....
> But just because we can *conceive* that functionalism is inadequate
> doesn't mean that it *is* inadequate.
> ....

Yes.

Moreover, people can easily fool themselves into thinking that they
can conceive of something when all that's happening is that they are
assembling words and phrases which happen to resonate with echoes of
meaning e.g. when they think the following are sensible questions:

    "What is the meaning of life?"
    "What time is it on the moon when it's midnight in New York?"
    "Are mathematical truths discovered or invented?"

> Second, "the apparently logically coherent possibility that *all*
> functional activity could go on without any experience occurring" only
> obtains if functionalism is false.  If qualia are identical to some set of
> conditions, for instance a certain set of up-and-running
> behavior-controlling functions, then it is not logically possible for
  ^^^^^^^^^^^^^^^^^^^^
> those conditions to exist and qualia not exist.

Right.

But note that some of the behaviour controlling functions may be
controlling "internal" behaviour, e.g. controlling the production and
modification of other behaviour controlling functions. I think this is
close to the position of Gilbert Ryle's 1949 book The Concept of Mind,
though most readers at the time misconstrued it as behaviourist.

Also note that specifying the relevant notion of "logical possibility"
needs some care. It's very easy for people to say "Because I can
understand the question 'Is X the same as Y?' it follows that the
concept of an X is different from the concept of a Y and therefore
there`s a logical possibility that X and Y are different. This notion of
logical possibility is very weak. (It was actually used around the turn
of the century by G.E.Moore to show that any definition of "goodness"
could be refuted.) What its proponents don't realise is that in this
sense it is logically possible for most theorems of mathematics to be
false.

Anyone who really wants to make progress with this sort of question
would do well to work on the *very very hard*, but real, problem of
trying to find out what sorts of designs really do have the ability to
support all the enormously varied functions of a human like brain (or a
chimp-like brain, or a magpie-like brain, etc.).

Currently we have very little idea of what these functions actually are
(including all the internal functions), and even less idea of what sorts
of architectures and mechanisms can support them.

So when people claim to be able to see a conceptual gap between
consciousness and explanatory mechanisms, all that's happening is that
they are unable to grasp conceptual relationships between something
ill-defined, and something they barely understand because it has not
been specified yet.

Of course they will then "experience" a gap! And of course it follows
that in the very weak sense of logically possible, it's a logical
possibility that there is a gap.

But nothing of any interest follows from that, for such gaps exist
between all pairs of concepts whose essential equivalence is not easy to
grasp.

When we have a good understanding of what can and what cannot be
done by various sorts of architectures we shall have a much clearer
understanding of what functionalism is actually claiming than any
functionalist can possibly have now.

I.e. functionalism is hardly a theory at present: it's a research
programme that's still in its very early infancy.

My own conjecture, as hinted at obscurely in my previously announced

    http://www.cs.bham.ac.uk/~axs/misc/like_to_be_a_rock

is that all the phenomena that make philosophers (amateur and
professional) want to talk about consciousness and qualia, will
*necessarily* be replicated in any architecture that has sufficient
richness of functionality to support typical human-like capabilities

    including the ability not only to see the table in front of you,
    which you presumably share with many other animals, but also the
    ability to pay attention to how the table looks to you (which may or
    may not be shared by other animals, and may or may not be shared
    with other robots, depending on how sophisticated their internal
    self-monitoring is).

It is very important that this is not just a matter of replicating (or
explaining) *external* behavioural capabilities.

Two engineers could build different machines with identical external
behaviour and quite different internal architectures, e.g. one driven
largely by pre-computed tables, the other computing what is needed as
and when it is needed. The former would not have the sort of
functionality I am talking about. (That's one way to build a zombie, if
you like: i.e. in that sense zombies are logically possible, contrary to
what Dennett writes, because it is logically possible to have the
shallow external functionality without the deep internal functionality
that characterises human mentality).

A corollary of this is that external behavioural tests can never
*suffice* to determine the existence of the sort of functionality
software engineers think about (deep functionality).

A further corollary is that if the conjecture about the identity between
(deep) functional processes and conscious phenomena turns out to be
correct, then there cannot be behavioural tests that are adequate (on
their own, without any information about the underlying implementation)
for determining which agents or organisms are conscious.

This conceptual gap between (externally observable) behaviour and mental
states is a consequence of a deep functionalist theory.

It isn't often noticed: on the contrary many people wrongly assume that
functionalism implies that there must be behavioural tests for mental
phenomena.

A shallow form of functionalism would entail that there are external
behavioural tests for intelligence (e.g. some form of the Turing test:
which Turing himself did not propose as a test or criterion for
consciousness or intelligence).

(Human sleepwalkers pass some behavioural tests for being conscious. Are
they?)

The common and tempting counter-argument that all we have as a basis for
judging one another to be conscious is behaviour, is, I think, invalid.

I don't judge people to be conscious simply because I check out their
behaviour, and probably neither does anyone else. We judge them to be
conscious because our brains are built (presumably as a result of
evolutionary processes that I don't understand) so that we have *no
choice* but to treat other people (and many animals) as sentient: i.e.
it is not a rational decision based on weighing up the evidence, but an
instinctive reaction, widely shared between all normal human beings. (I
am not so sure about autistic individuals.)

Why this should be so is a topic for another occasion. (Should a mother
whose infant screams or gurgles and smiles wait for convincing evidence
of consciousness before being committed to a belief that that child is a
person like her? A philosopher might say "yes". But a biologist could
give a different answer. Likewise an engineer trying to design an
effective social and reproductive system.)

If you make certain assumptions about classes of implementation engines
available, then you can argue that external behaviour rationally
supports some theories about what's going on inside, because using those
engines certain implementation strategies could not work. E.g. given
assumptions about available storage capabilities you can show that the
look-up tables required for precomputed zombie-type implementations
would be far to big to fit into a brain (or even into our universe). But
that's not arguing solely from behaviour. (Also it's not what we
naturally do, since we naturally do not base these beliefs on argument
or evidence at all. How many parents know anything about storage
requirements for zombies?)

> ....
> ...The "first-person"
> ends up being something "third-person", something out there in the world.

Nicely put. I completely agree.

When people argue that that must be wrong because there's a conceptual
gap between the two, we then have the task of convincing them that the
gap is a subjective aspect of THEIR impoverished grasp of the problem,
not a real conceptual gap.

That's not easy to do. I've succeeded only in a small percentage of
cases where I've tried. The confusions are very persistent, very
compelling, very deep. (It took me at least ten years to change, between
the days when I was a research student, firmly convinced of the gap, and
the time I gradually came to realise I had no basis for the conviction,
mainly as a result of much discussion about how human-like robots might
be designed.)

The difficulty of convincing people that the gap is illusory is no
accident: the same problem will occur with robots as soon as they have
sufficiently sophisticated functionality, including useful internal
self-monitoring and self-reflective capabilities.

For these will always be only partial, and will not give the poor robots
a complete or completely accurate account of what's happening inside
them. Having the impression of a cartesian theatre is, I think, a
consequence of design requirements for a fully functional intelligent
agent: e.g. one who can plan the design of a shelter by thinking about
how easy it is to see it from outside.

"If I put these branches here then until I come very close I cannot make
out the entrance" is a comment about what's in the cartesian theatre,
and not just about what's out there. Thus both the robot home-maker
who is afraid of predators and the designer who wishes to impress others
with the quality of the family residence will need access to what
strikes them as an internal theatre. And theories of immaterial qualia,
etc. will inevitably be born again.

Isaac Asimov noticed this long ago, in his story about robots who became
cartesian dualists, invented religion, thought their souls could survive
the destruction of their bodies, etc. etc.

And of course many such non-zombie robots would write to psyche-D, using
conceptual gaps, possible global colour switches, etc. to argue
earnestly against materialism, functionalism, computationalism, etc.
(Those who take up robot-design as a serious study might come to see
their confusions.)

> ....
> So when Stapp says
>
> "...there is something very unnatural about certain functional structures
> being also experiential structures when this fact makes no dynamical
> difference: the experiential aspect, which must be logically
> distinguishable from the functional aspect if the latter could conceivably
> occur without the other, seems superfluous."
>
> he presumes the non-identity of experience and function, which of course
> makes it seem that experience would make no "dynamical difference" (be
  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> epiphenomenal) since one could conceivably occur without the other.  But
> if experience turns out to be *in fact * identical to function, then
> experience *does* make a dynamical difference, precisely the dynamical
  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> difference that the function makes.

Right. And we cannot even discuss such questions sensibly without a
proper understanding of the full range of relevant dynamical
differences, i.e. the full range of functionality required to support
the full range of human-like capabilities and experiences. I've hinted
above at how attention to qualia can make a big difference to one's
ability to build a safe shelter. That's certainly a functional role for
qualia.

I think most or all of the people (e.g. Henry Stapp?) who say
experiences cannot play that role have no idea what the required roles
are. If they did, they would be able to specify how to design robots
functionally equivalent to ourselves. Nobody is anywhere near this yet.

(Nor anywhere near a specification of which subsets of these functions
might exist in various different architectures implementing different,
more or less human-like minds, e.g. various kinds of animal minds, human
neonate minds, Altzheimer minds, schizophrenic minds, etc.)

We won't have that understanding without a good theory of what sorts of
architectures have which sorts of capabilities.


I don't mean just the physical or neurophysiological architecture.

As many software engineers take for granted, but apparently some
scientists and philosophers don't, you can have layers of functional
architecture at different levels of abstraction in a hierarchy of
implementation levels: all of them real, all of them causally
efficacious: a change in one lisp or prolog datastructure can cause a
change to be made to another one. A change in one desire or belief can
cause changes to occur in others. A change in the level of poverty can
cause a change in the amount of crime. All of these illustrate causation
in virtual (abstract) machines. The causal roles of qualia are just
another case.

In all cases the virtual machine is *implemented* in physical
mechanisms, but it's a different machine with its own ontology and its
own laws, neither expressible in the language of physics (or
neuroscience, for that matter.)


It remains an open question which kinds of layered architecture could
have layers corresponding to human experiences of which we are aware
along with layers containing the vast iceberg of mental processes of
which we are totally unaware, such as the syntactic processes involved
in understanding or generating English, or the learning processes that
go on when you improve your touch-typing or violin-playing with
practice, or the concept-formation processes that go on when a child
develops a concept of infinity or an adult learns category theory for
the first time.

An open question is to what extent coupling between layers is a
requirement for that functionality (which is one way in which quantum
mechanics might prove relevant, since it seems to allow close coupling
between large scale and small scale phenomena to change the dynamics of
the small scale -- or that's how it seems to me, a complete amateur).

    (Could this turn out to be partly analogous to the ways in which a
    machine with writeable control store can allow high level software
    to change the microcode at run time, or some architectures allow low
    level interrupts to invoke high level interrupt-handling software
    which then re-directs the normal sub-machine-code processes? Here we
    have very useful kinds of coupling between processes in very low
    level and very high level virtual machines, even though mostly
    a good designer will keep the levels separate.)

If someone objects that they can TELL by introspection that their
experiences just exist without any causal connections, why should we
believe him?

Turning Hume on his head we can say: you cannot *experience* an absence
of causation, any more than you can experience causal connections.

>...
> If and when plausible identity conditions for qualitative experience are
> empirically discovered, then we won't any longer be in a position to ask
  ^^^^^^^^^^^^^^^^^^^^^^
> the "hard" question, "Why do qualia arise from these conditions?" since
> they aren't arising from the conditions, they simply *are* the
> conditions.

Yes. Nicely put.

But "empirically discovered" is not the right phrase. It will require
considerable conceptual creativity, like the discovery of the deep
isomorphisms (identity?) between aspects of geometry and of arithmetic.
E.g. the very same mathematical notion of group is applicable to sets of
geometric transformations and to sets of arithmetic transformations; and
the very same notions of continuity are relevant.

If it were merely an empirical discovery the conceptual gap would be
real, and the discovery would be a shallow, fragile thing, liable to be
upset by new data.

> ..If, perversely, we don't accept this point, then there can
> never be closure on the qualia chase since no matter how precisely one
> specifies the conditions under which qualia occur, it can always be
> objected, "Well, all you've done is show a *correlation* between the
> conditions and qualia."

There will always be such objectors. But it is up to the objector to say
clearly what these alleged qualia are that "merely correlate" with the
functional conditions.

In my experience of arguments with such objectors, at that point all
they can do is invite one to engage in some sort of introspection to see
what is being talked about.

For me introspection reveals nothing that could exist independently of
functional relations.

E.g. How things LOOK to me, to choose one example of a class of qualia,
far from being non-functional, is a highly functional aspect of what I
am able to do, what I can compare with what, what actions I can think of
as relevant, how I need to move to see more or less of something, etc.

(I think Wittengstein saw this in his typically obscure way when he
wrote that having an experience, e.g. "seeing as", necessarily involves
"mastery of techniques", or some such thing, in his Philosophical
Investigations. Ryle was clearer when he characterised all mental states
as inherently dispositional. Maybe he got the idea from Wittgeinstein
and tarted it up a bit?)

Similarly, experiencing pleasure or pain is *inherently* bound up with
control processes to do with making it more or less likely that you'll
try to maintain (or increase) whatever is currently going on or whether
you'll try to stop or reduce it. Besides these short term control
functions pleasure and pain can have longer term causal links through
learning mechanisms. (Of course in some cases superimposed functional
states can produce conflicting control states, e.g. masochistic sex,
where the pain and the pleasure are deeply intertwined: but that's
because conflicting dispositions, conflicting deep functional relations,
can coexist in a sufficiently rich functional architecture. Ask any
dsigner of operating systems.)

I haven't a clue what people are talking about when they say they have
pains and pleasures or experiences which they can conceive of as totally
dissociated from functional relationships. I suspect they just haven't
learnt to pay attention properly (like people who sometimes don't notice
when they are angry, or in love).

One counter-argument I've met is that such functional states and
processes are involved in the many control processes that go on in
brains of which we are not aware, and therefore having those functional
roles cannot suffice to make them conscious states and processes. There
are at least two replies to this:

(a) It ignores the possibility that identical functional processes may
occur sometimes in the context of additional self-monitoring mechanisms
sometimes not. In the latter case their functionality will be enhanced,
e.g. because they can contribute to types of reflective analysis leading
to certain kinds of learning or finer control. (I.e. functionalty is
composite, not atomic.)

(b) It ignores the possibility that some of the processes in ourselves
of which we are not aware are exactly like the processes of which we are
aware, except that WE cannot access them directly. They may be accessed
and monitored by other processes within us that we don't know about,
just as states in your mind are accessible to mechanisms that I don't
know about but not to me.

I.e. the counter-argument implicitly (and unwittingly) links
consciousness to the abilities of a *particular subset* of control
mechanisms (the "highest-level" control mechanisms, which have
external-reporting capabilities, as Dennett once suggested ???).

That's a conceptual restriction on the notion of consciousness that
could be a powerful conceptual blinker, ruling out the possibility of
forms of consciousness in ourselves of which we are unconscious!!

By unblinking ourselves we may be able to deal a lot better with the
paradoxical cases, like the sound that you only notice when it stops,
the experiences had in dreams when you are unconscious, the
perceptual capability of a sleepwalker who is unconscious yet dresses
himself and opens the door, and the grammatical analysis of complex
sentences that contributes to the experience of understanding what is
said, for instance when reading long and involved sentences like this
one.

(From this standpoint there is absolutely nothing to be surprised about
in the experiments demonstrating blindsight. I've never understood the
fuss they caused. Similarly multiple personality disorders. These and
other phenomena are only to be expected in a rich enough architecture,
admitting many forms of malfunction or diversity of detailed
functionality: like the differences between Mozart and me.)

The history of science and mathematics is full of cases where conceptual
blinkers had to be removed. E.g. treating 0 as a number, thinking about
influence at a distance as physical influence, grasping that continuity
and differentiability could come apart, abandoning the idea of a unique
space-time framework, grasping the possibility of space-filling curves
(which were *obviously* impossible, until their possibility was proved),
accepting wave-particle duality as not a contradiction, allowing
evolutionary processes involving no designer to create sophisticated
designs, etc. etc.

So come on folks: unblinker yourself and let yourself contemplate the
existence of lots of experiences within you of which YOU are
unconscious. (Which is not to say that NOTHING is conscious of them:
they might be monitored and used by sub-processes about which you know
nothing. What's your cerebellum doing right now?)

Some of the people who think there's a conceptual gap between functional
states and experiences also think that talking about the intelligibility
of flipping qualia (colour experiences being swapped, etc.) adds support
for their claims.

But I don't see how it does any more than say that it is logically
possible that some sort of gigantic flip could occur that is essentially
like the flip that ALREADY occurs when we look at a necker cube. There's
no reason to suppose that THAT sort of flip is functionally
disconnected, or that it is possible for any such flip on a more global
scale to occur that is functionally disconnected. (Bernard Harrison once
wrote a book arguing that case. I think it may have been called
something like Form and Content. It's probably at least 25 years old,
maybe more.)

When people try to add further specification to the claim that a
functionally disconnected giant colour-qualia-flip could occur, I think
they unwittingly end up with something incoherent, just like a
philosophical sceptic trying to add further specification to the idea
that everything in the universe might be moving left at three miles per
hour without the movement being detectable. But incoherence never
emerges because they usually never complete the description, they just
assume it's obvious that it could be completed in a coherent way. Is
it???

("But I can conceive of it" is no answer: it might just report a
conceptual delusion, of which there are many.)

>
> If we do accept it, then the big question is, how do we know that we've
                                                ^^^^^^^^^^^^^^
> found the state of affairs that genuinely constitutes experience?

I agree. That's an important question, whose answer is far from obvious,
and which should get more attention.

Maybe we won't be able to answer it till we know a lot more, just as I
suspect Aristotle could not possibly have given a criterion for adequacy
of a physical theory that would have included things like Maxwell's
electromagnetic theory, relativity theory, or quantum theory.

Until we have our theory of deep functionality we may not have a good
basis for saying why some of the things it refers to are identical with
what we previously called experiences, or consciousness.

I.e. we should just get on with the research.

> We will
> know that we have the identity conditions in hand when the simplest, most
> ontologically economical, empirically predictive, and heirarchically
> integrated theory emerges out the data, not when our favorite picture of
> what consciousness simply *has* to be is vindicated.

Although I agree with the gist of this, I think it is going to be more
difficult than this suggests.

First I don't think any theory will ever "emerge out of the data":
psychologists, neuroscientists, biologists, sociologists, linguists,
anthropologists etc. can go on collecting data forever without the sort
of theory we need ever emerging.

The theory will have to come, in part, from a highly creative
theoretical (i.e. non-empirical) process of exploring the properties of
ever more sophisticated designs for functional architectures, done in
parallel with empirical studies. The empirical studies will probably
have to be partially directed by questions emerging from studies of
possible architectures.

Empirical data will feed back suggestions for the explorers in design
space (e.g. showing them constraints in available mechanisms, and also
giving them ideas about types of mechanisms they would never have been
able to think up in their armchairs).

We may need new types of mathematics to characterise the properties of
such designs and their functionality. (Some people think they are
producing the relevant mathematics in the study of dynamical systems,
including chaos, attractors, etc. I am suspicious that that sort of
thing won't suffice, for reasons I have given elsewhere, but it's worth
a try, alongside other approaches. We need open minds.)

> If we insist on a
        ^^^^^^
> particular view of what the identity conditions of consciousness must be
> (a special aspect of as yet to be discovered informational
> something-or-other,

I agree we should not *insist* on this (apriori insistence has no place
in science).

However some people (not all, but more than now) should work on the
*conjecture* that when we understand how information processing is not
just a matter of syntactic transformations but essentially involves
*control* (including control of information processing) and, in some
cases *semantics* (including information structures within an
architecture that refer to other information structures in the
architecture), then we'll get some new deep conceptual insights.

In particular, we'll make a type of progress that cannot come by other
routes.

At a later stage, the important gaps in knowledge will lie elsewhere.
I obviously can't say what such gaps are.

Are quantum mechanisms going to be relevant? Maybe, but none of the
arguments so far produced are valid: they seem to be full of category
mistakes, for a start.

Here's a hint of a different sort of argument for the necessity of
quantum mechanisms underlying intelligence: we need quantum mechanical
systems in order to provide large enough, stable enough, intricate
enough, compact enough, information stores. (I suspect classical
mechanisms could not work for brain-sized memories with human
information richness because the components would have to be so small
that they would be subject to thermal buffeting, which would cause too
much information to be lost over time. Quantum mechanisms seem to this
layman to be required for the long-term stability required for
preservation of intricate sub-microscopic structure. That's not a new
idea: it's often said that without quantum mechanisms genetic
information could not be preserved across generations. NB this has
nothing to do with Goedel's theorem, and it applies as much to compact
computers with large information stores as to brains.)

> ..But to repeat, when a
> successful theory converges upon identity conditions, there is no further
> question about explaining why qualia arise, since you've *got* the
> slippery devils.

Yes.

But there will always be people who say: No.

For they will look at the theory, and they will look at their own
experiences, and they will say "There's a conceptual gap between these
two". And nothing we can say will remove this tendency.

In some cases it will be due to metaphysical neurosis (i.e. wanting
human beings to be something special, and hating the thought that we
could essentially be built in factories).

In some cases it will be because of prior commitments to religious
beliefs which rule out this kind of view of minds (or souls).

But there will always be a subset who, without being subject to
irrational pressures, just cannot grasp the identity, just as there are
people who cannot see that it is possible for something to be
simultaneously decreasing its velocity and increasing its acceleration
(or vice versa) or who cannot understand complex mathematical proofs.
(E.g. I am sure I'll never understand the recent proof by Wiles of
Fermat's last theorem.)

I.e. the kind of identity in question is not a simple "obvious" identity
but a deep and complex one: there's one thing that can be looked at in
two very different ways. (Like the geometrical and the set-theoretic
views of the real number continuum: not everyone who understands some
geometry and some arithmetic can grasp the underlying identity.)

It seems to me, alas, that our evolutionary and educational processes
seem to produce a subset of people who lack the capability to grasp some
of the complex relationships that other people can grasp. But it doesn't
stop them asking the questions: they just can't understand the answers.

It's very sad. (I feel sad about the ones I can't grasp. I'd LOVE to
understand why there cannot be an integer N > 2 and three non-zero
integers a, b and c such that a^N + b^N = c^N. But a recent TV program
on Fermat's theorem convinced me that I would never understand all the
relevant details, unless I spent the rest of my life working on nothing
else, and probably not even then. Similarly I feel sad about people who
will never grasp the potential of intricate information processing
architectures to replicate everything there is to replicate about mental
states and processes.)


Maybe when we have a good theory of the mechanisms underlying deep
mathematical and scientific understanding we shall be in a position to
help the people for whom that is currently impossible, for reasons that
we now don't understand. I.e. maybe we'll find good ways to extend
people's mental capabilities? Maybe not.

>
> The issue of how we'll know when consciousness is explained brings up
> Torfi Sigurdsson's excellent point some time back that we not only have to
> conduct research, but that we must also be willing to question and adjust
> our "first person" notion of consciousness in the light of that research.

Yes. We have to be willing in principle to question anything and
everything. (Except that.)

> Such conceptual change relating to the explanandum has happened repeatedly
> in the course of scientific discovery (e.g., the concepts of space,
> matter, life) and there is nothing privileged or unimpeachable, I daresay,
> about anyone's conception of qualitative consciousness.

And the history of the psychology, sociology and politics of science and
mathematics shows that there can always be strong pockets of resistance
that have nothing to do with whether the conceptual change is in the
right direction or not.

>
> Certain solutions to the mind/body problem are ruled out a priori by some
> theorists simply because of their picture of experience itself.

I would say it's the shallowness of their picture of experience, that
rules it out.

> ..Could
> one's theoretical commitments be driven by one's conception of experience,
> perhaps, and if so, how does one justify that conception?  For instance,
> many think it obviously the case that qualia have a non-functional
> component or aspect, but that point has to be argued, not assumed.

Someone who finds the gap self-evident simply because he lacks the
capability to grasp how such an identity could possibly exist, will
never be able to produce an argument. So don't be surprised if there are
always people who go on producing repeated assertions, but no arguments.

This is inevitable because of how some minds work. I fear that
discussion with such people will actually get nowhere.

People who share the views and conjectures I've outlined should stop
wasting their time arguing and instead get on and do the research that
needs to be done, and then maybe in 10, or 50 or 500 years time report
back showing in some detail how the identity works. (Records of many
contemporary discussions of consciousness will then appear very silly.)

Perhaps the combination of fully worked out theory and working
demonstrations (robots built on the basis of the theory and intricate
non-invasive ways of observing and manipulating their and our mental
states) will help to convince the remaining doubters, in a way that "in
principle" arguments cannot.

I doubt it will convince everyone.

There will always be doubters. Even some of the robots will be doubters,
because of the way their ability to have experiences has been
implemented, giving them only a very shallow view of what's going on
inside them, just like most contemporary contributors to discussions of
consciousness.

> --Tom Clark (twc@world.std.com)

Aaron Sloman, ( http://www.cs.bham.ac.uk/~axs )
School of Computer Science, The University of Birmingham, B15 2TT, England
EMAIL   A.Sloman@cs.bham.ac.uk
Phone: +44-121-414-4775 (Sec 3711)       Fax:   +44-121-414-4281
==========