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

This file contains three messages posted to psyche-d (which
is relayed to sci.psychology.consciousness.

First is a message from Henry Stapp to me attempting
to characterise my views and comment on them. This included
a message that he had posted to psyche-d and then withdrew,
along with an accompanying note to me.

Next is my reply to the points he made in the accompanying
note.

After that is my reply to the points made in the message
he withdrew.

Subsequently he agreed that the whole correspondence should
be posted.

Note: psyche-d discussions are archived at
        http://www.ai.sri.com/~connolly/psyche-list/


From STAPP@kelvin.lbl.gov Thu Feb 13 17:00 GMT 1997
Date: Thu, 13 Feb 1997 7:47:11 -0800
From: STAPP@theorm.lbl.gov
To: A.SLOMAN@CS.BHAM.AC.UK
Cc: STAPP@theorm.lbl.gov
Message-Id: <970213074711.2a80159d@theorm.lbl.gov>
Subject: Re: Qualia & Form Perception

From Aaron Sloman Sat Feb 22 09:45:22 GMT 1997
To: psyche-d@rfmh.org
Subject: Qualia & Form Perception

Dear Psyche-d readers,
On Thursday 13th Feb, Henry Stapp sent me the message below.

A few days later I responded to him, then he agreed that the
whole correspondence should be posted to psyche-d. I shall send
my reply after sending this, and I apologise for any overlap with
discussions that have occurred in the mean time.
Aaron
=======================================================================
From: STAPP@theorm.lbl.gov
Date: Thu, 13 Feb 1997 7:47:11 -0800
Dear Aaron;

After a long absence from psyche-d I did read your Feb10 posting
and sent a reply that I will append to this one. However, upon
reconsideration I had Patrick kill it, as I thought that I had
completely misunderstood your position.

But let me now state to you privately what I take your position
be, and state my objection to it.

You do not deny the existence of consciousness, but rather
claim that it IS a certain functional activity of the brain/body.

Take a weather system of hurricanes and storms etc. One can
describe it in terms of motions of atoms and molecules etc, but
a more useful description for some purposes would be in terms
of high pressure and low pressure areas and other gross features
that could be the basis of a high-level causal description.
Your idea is that the description of  brain activity in
terms of one's consciousness experience is, at least
in first approximation, rather like this high-level
description of a weather system. Am I close?

If so, then my objection is this: I do not object at all if you
do not try to maintain also that the lowest-level description
(or an approximation to the lowest-evel description that is
completely adequate for the discussion of the mind-body problem)
is in terms of the concepts of classical physics.
But if you hold that one can take  a classical physical
description to be in principle, in this context, completely
adequate for descibing a mind/body system at the micro level,
then my  objection to the analogy is that the classical-physics
description is, within the classical physics framework complete:
every aspect of the weather system is in principle deducible
from the micro-description in terms of the motions of atoms,
,ions, molecules, electrons, and electromagnetic fields,
and the laws and principles of classical mechanics provide in
principle the foundation for a complete understanding of the
causal relationships in the high-level complexes. But one cannot
DEDUCE the presence of the experiences of pain, or saddness
or joy from the principles of classical physics and the micro-
description: one can POSTULATE certain connections between the
classical-physics description and experiences, but cannot deduce
them from the principles of classical physics and the
micro-description in terms of the local elements of the
classical-physics description. Thus the existence and form of
conscious experience is not a logically necessary concomitant
to the classical-mechanics description of a body/brain in which
conscious experience is present, whereas the existence and form
of a hurricane is entailed by the classical-mechanics description
of a  situation in which a hurricane is present

For this reason I regard the classical-mechanics description of nature
to be deficient in the context of a study of the mind/brain problem:
a good theory of closely related phenomena should hang together better
than that.


Best regards, Henry

I append my recalled posting to psyche-d:

Aaron Sloman's tack (Feb 10)seems to be to admit into his realm of discourse
about what we would normally call " one's conscious experience" all of the
functional properties that go along with one's conscious experience, such as
memory/recall, global access, pictorial or linguistic representability, etc.
but to ban or banish the idea of "my conscious experience" which for
most of us is the core reality that seems to lie at the center of this
functionality.

Since we are committed to a scientific approach Aaron's approach seems hard
to fault, because it promises to be able to deal with all of the
functionality associated with one's conscious experiences without embroiling
us  in the difficulties associated with the attempt to bring the concept of
consciousness itself into scientific theory.

Still, admitting the concept of one's conscious experience into the realm of
theoretical discourse does bring certain theoretical advantages. For one
thing, it provides clues to the functioning of our brains.
There are, ab initio, many possible ways that one might invent to try to
account for human gross behaviour patterns. The information provided by
the existence and structure of one's own experience can provide important
clues as to how the internal processing is organized: there must be in an
adequate account of brain processing the description of  organizational
structures that can account for one's "seemings": for what one seems to
experience. Thus these "seemings" serve as  clues to
the organization of brain processing.

But these "seemings" are more than mere clues. Explaining these seemings
is  necessary  for a satisfactory theory. For within web of
functionally related theoretical structures one must in the end identify
the realities against which to test the theory. These can include observable
gross patterns of behavior, and a rather satisfactory theory might
be one that merely aims to explain these observable gross patterns of behavior
in terms of internal processing structures. But there are in nature, not only
these observable gross patterns of behavior, and the internal processes that
occur on our brains and bodies but that are not "seemings" ( i.e., are not
conscious experiences): there are also these "seemings". Now it may well be
true that a theory of internal process can be constructed quite well without
initially paying any special attention to these special effects, or spelling
out how they are related to the vast internal process within which they
are imbedded: it may well be that the most expeditious way to proceed with
the construction of a theory of internal process is to ignore these seemings
in the initial phase of construction of the theory. But in the end these
seemings are part of reality and, presumeably, part of the internal process.
They are part of what must be explained by a complete theory of the internal
process. Yet they are certainly not identifiable with the whole of the
internal process. So there is a bona fide problem of explaining how they are
related to the whole internal process.

It is in fact not knowable before the construction of an adequate theory
of the internal process whether it will be best to first ignore these
"seemings", and incorporate them later as special and perhaps
bizarre side effect, or whether it will be best to take them as essential
features that lie at the very core of our high-level internal processing.

Plausibility arguments can be given for each of these approaches. In the
end some people's intuitions on this methodological issue will be better
than those of others. That's how science progresses.

Henry P. Stapp
http://www-physics.lbl.gov/~stapp/stappfiles.html


From Aaron Sloman Fri Feb 14 12:05:51 GMT 1997
To: STAPP@theorm.lbl.gov

From Aaron Sloman Sat Feb 22 10:58:52 GMT 1997
To: psyche-d@rfmh.org
Subject: Re: Qualia & Form Perception (reply to Stapp)
Cc: STAPP@theorm.lbl.gov

This was my first reply (sent 14th Feb) to Henry Stapp's message
of 13th Feb. After sending this one I sent a shorter summary of
points of agreement and disagreement, which I'll post after this.

In this message I've corrected some typos, changed the format to
better fit the html psyche-d archive, and added a few explanatory
notes all starting "[Note added" and ending "]".

===
Dear Henry,
Thanks for your useful comments.

[Henry]
> You do not deny the existence of consciousness, but rather
> claim that it IS a certain functional activity of the
> brain/body.

Not exactly.

I claim there is no single IT that is consciousness. Rather our
ordinary concept actually refers to a mish-mash of different more
or less closely related phenomena which do not all necessarily
occur together (e.g. different subsets occur in different
organisms and machines).

But yes, as far as I can tell at present, each of those phenomena
is, or is an aspect of, some functional activity or activies
implemented in the brain/body (the combination is important,
since some of them involve fairly tight control loops including
physical devices like eyes and fingers).

[Note added 22nd Feb: some reactive mechanisms may be composed
entirely of continuous feedback loops. However, deliberative
mechanisms (as I've defined them) inherently involve the
*additional* ability to assemble complex structures in discrete
steps, and to use associations between discrete structures. I.e.
these deliberative mechanisms are inherently digital. When the
two are combined, analog to digital and digital to analog
converters are required. This acknowledges points made by other
posters in the last few days about the role of feedback loops. I
find it useful to think of a mind as essentially a control
system, though early control theory used too small a range of
mechanisms, as Norbert Wiener acknowledged in the second (1961)
edition of his 1948 book.]

[Henry]
> Take a weather system of hurricanes and storms etc. One can
> describe it in terms of motions of atoms and molecules etc, but
> a more useful description for some purposes would be in terms
> of high pressure and low pressure areas and other gross features
> that could be the basis of a high-level causal description.

I am not simply concerned with what is useful for external
observers and predictors, but with deeper features that
distinguish different sorts systems that we find in our world.

There's a big difference hurricanes and animals like fleas or
monkeys.

In living systems the most important thing about them which
explains what they are, how they came to exist and how they
behave and develop is the fact that they have, acquire, store,
modify, transform and use information, both in keeping themselves
and others alive and also in reproducing.

[Note added: and of course doing many other things which serve
these ends, and some which don't, but may have hitched a ride on
the goal achieving mechanism.]

I think this is especially true of mobile animals. I think it is
also true of plants, but to a much lesser extent. I am not sure I
fully understand the difference.

[Note added 22nd Feb. I've recently learnt about infected plants
"communicating" with other plants by releasing a gas that "warns"
uninfected plants to start preparing their defences: a very
primitive sort of communication of information.]

By contrast, hurricanes are not essentially information
processing systems, as far as I know, though they are energy
transformers and they impact on their environment. These are two
properties they share with animals.

They also have a certain coherence and stability (at least for a
time), as does the planetary system (for a longer time), and this
may make them appear essentially similar to animals. But I think
this is a superficial similarity, since they are not information
processing systems.

E.g. I don't think hurricanes and planetary systems ever make use
of any specification of a future state to be achieved or of a
remote object that has been sensed as something to eat, mate
with, escape from, inspect to acquire more information, etc. The
only influence that a remote object can have on a hurricane is
energy transfer. (Of course energy transfer is also involved in
sensing, but that's just an implementation detail).

[Note added: if it turned out that hurricanes could select routes
that enable them to reach locations where they'd receive an
energy boost allowing them to survive longer, then I'd have
to change my views about them.]

[Henry]
> Your idea is that the description of  brain activity in
> terms of one's consciousness experience is, at least
> in first approximation, rather like this high-level
                          ^^^^^^^^^^^
> description of a weather system. Am I close?

Yes: though I would not express it like that because I'd like to
stress the differences as well as the similarities.

[Note added: See my response to Pay Hayes with subject line
"A fourth type of answer". Talk about differences of degree, or
"rudimentary" types of consciousness can be misleading, if the
features that are missing are not explicitly indicated. The space
of possible mechanisms we are discussing is not a seamless
continuum.]

[Henry]
> If so, then my objection is this: I do not object at all if you
> do not try to maintain also that the lowest-level description
> (or an approximation to the lowest-evel description that is
> completely adequate for the discussion of the mind-body problem)
> is in terms of the concepts of classical physics.

Actually I have a completely open mind on this. After all I
believe that at present the lowest level description that's
scientifically justified must be quantum mechanical: that's how
things are.

Maybe something deeper will emerge in physics in 100 or 800 years
time?

But all that is equally true of computers, hurricanes, digestive
processes, etc.

What's not clear is which features of quantum mechanics make a
difference.

I have objections to the arguments which purport to make a link
between the quantum mechanics and so-called consciousness. My
problem is that they all start either from

    (a) a specification of "consciousness" which I do not
    recognize at all, or find incredibly vague,

or from

    (b) a specification which describes only a tiny subset of the
    abilities of a small subset of animals (e.g. the ability of
    some humans to understand G\"odel's incompleteness theorem).

or they make inference leaps that I find incomprehensible between
the different levels of description.

[Henry]
> But if you hold that one can take  a classical physical
> description to be in principle, in this context, completely
> adequate for descibing a mind/body system at the micro level,

Actually I don't believe that classical physics is adequate to
describe how computers work in detail. But maybe that's
irrelevant for now.

> then my  objection to the analogy is that the classical-physics
> description is, within the classical physics framework complete:
> every aspect of the weather system is in principle deducible
> from the micro-description in terms of the motions of atoms,
> ,ions, molecules, electrons, and electromagnetic fields,

I presume this is true of many of the high level concepts used to
describe hurricanes, though I don't know whether the butterfly
effect is capable of enabling non-classical phenomena to produce
global effects that could not have been predicted classically. I
suspect your arguments are not affected by that.

However when hurricanes are described as "dangerous" "causing
damage", or "economically disastrous", new concepts are
introduced that depend on higher level systems that cannot be
defined in physical terms and whose properties cannot be deduced
by logic from those of physics (though they are implemented in
the physical world). That's because they use socio-economic
concepts, which refer to a different level of reality from
physics.

[Note added: these terms are "relational". They describe
relations between elements in a complex causal network.
Similarly, the terms we use to describe mental states and
processes, e.g. "believing", "noticing", "considering",
"selecting", "enjoying", are relational in that they describe
relations between elements in a cojmplex causal network, which
nobody so far has characterised adequately. In both cases
describing the causal networks requires a vocabulary that cannot
be defined in terms of the vocabulary of physics. But in both
cases the networks are *implemented* in physical systems.
Incidentally I suspect the variety of types of networks that are
entirely continuous is far more limited than those that include
discrete/digital sub-mechanisms, and evolution "discovered"
that and put it to good use.]

> ...

[Henry]
> But one cannot
> DEDUCE the presence of the experiences of pain, or saddness
> or joy from the principles of classical physics and the micro-
> description: one can POSTULATE certain connections between the
> classical-physics description and experiences, but cannot deduce
> them from the principles of classical physics and the
> micro-description in terms of the local elements of the
> classical-physics description.

For me that's the wrong question. I don't think most of the
interesting properties of things implemented in the physical
world can be DEDUCED from the principles of any sort of physics.

[Note added: because the implicitly relational concepts used in
describing different sorts of causal networks cannot all be
defined in terms of physical concepts.]

If you and I have a game of chess, then no matter which medium we
use for it our game is inevitably implemented in physical
systems. However, I do not believe that any logical deduction,
starting from a *purely physical* description of that system and
its history, together with *laws of physics*, and nothing else,
can ever have as its conclusion something like:

    White (Stapp) checkmated black(Sloman) in 25 moves.

[Henry]
> Thus the existence and form of
> conscious experience is not a logically necessary concomitant
> to the classical-mechanics description of a body/brain in which
> conscious experience is present,

nor a logically necessary concomitant to any other physical
description, I would add.

I.e. I don't see any difference between classical and quantum
mechanics here. (Though you have tried to make me see it! Perhaps
I just lack the intelligence to understand???)

> ..whereas the existence and form
> of a hurricane is entailed by the classical-mechanics description
> of a  situation in which a hurricane is present

OK, subject to the various qualifications above.

> For this reason I regard the classical-mechanics description of nature
> to be deficient in the context of a study of the mind/brain problem:
> a good theory of closely related phenomena should hang together better
> than that.

I agree with you (and Jeffrey Gray, incidentally) that we need to
have a deep understanding of relationships between

    (A) "high level" systems (abstract virtual machines), which
    include internal and external causal relationships

and

    (B) lower level systems, e.g. the underlying physical reality,
    and various intermediate level virtual machines where
    appropriate.

This requires understanding how (A) can be implemented in and
supervenient on (B).

[Note added: lots of philosophers have tried to characterise the
possible forms of such relationships, e.g. Bill Robinson makes an
interesting attempt in Mind 1990, among others. But unfortunately
they tend to start from simple mechanical systems (e.g. alarm
clocks) which are fairly easy to characterise, and then jump
straight on to the much harder problem of the relation between
mind and brain. If only they'd stop to sharpen their tools on
compilers, wordprocessors, spreadsheets, operating systems,
computer networks, plant control systems (roughly in order of
increasing complexity) etc. then maybe they could produce
something more relevant to the task.]

I also agree with your (and Jeffrey Gray's) implicit claim that
this deep understanding [of the relationship between levels] will
not simply amount to the discovery of empirical correlations:
that's why I find all this talk about "the neural correlates of
consciousness" very unsatisfactory.

It is explanatory underpinnings we want, or implementation
theories, not just empirical correlations, which are what you get
from shallow experimental science.

[Note added: "shallow" does not mean "easy" or "worthless".
Shallow science is often an essential precursor to the deep
advances. Shallow science paints the dot picture that the deep
reconceptualisations allow us to view in a new way.]

However, this does not mean that the deep connections are
deductions.

The discussions are complicated by the fact that we don't yet
have any clear idea regarding what sort of (A) we are talking
about, though some people are convinced that they do have a clear
idea and I argue that they are just unwittingly deceiving
themselves.

Maybe we could make progress if we spent more time trying to
characterise what sort of (A) we are talking about, and trying to
clarify what sorts of explanatory relationships other than
deducibility and mere correlation can exist between different
levels of reality.

(Or, if you prefer de dicto mode, between levels of
description.)

I hope to read your original draft later. I have to rush now. If
you think it worth while we could post this or some modified
version to psyche-D.

I don't suppose you'll be at the Elsinore workshop in August? I
hope we do get a chance to meet and discuss things face to face
some time.

[Further comments to follow]
Aaron


From Aaron Sloman Sat Feb 22 11:11:39 GMT 1997
To: psyche-d@rfmh.org
Cc: STAPP@theorm.lbl.gov
Subject: Re: Qualia & Form Perception

This is my second response to Henry Stapp, sent to him on 13th
Feb, after which he agreed that the whole correspondence should
be posted to psyche-d. As before, there are some notes added, in
square brackets.
====

Dear Henry,

I've now had a first quick look at the message you posted and
recalled, and I will confirm what you presumably concluded
yourself namely:

I do not deny that there are ways things seem to us.

I agree with you that there are mental states and processes that
cannot be defined EITHER in terms of externally observable
relations between environment and behaviour OR in terms of
physical brain processes. I.e. they constitute a level of reality
that is not deductively reducible, just like poverty, crime,
ownership of houses, and kings checked by bishops in chess.

I also agree that we have internal access to a subset of those
mental states and processes in ourselves.

I also agree that there's a special subset of mental states which
can be described as how things seem to us, about which we cannot
be mistaken (at the time) since it is true by definition that how
things seem to us is how things seem to us. I.e. we cannot be
wrong about how things seem to us.

But that's a trivial logical truth that is easily mistaken for
something profound.

I am even prepared to admit that something like this is true:

> admitting the concept of one's conscious experience into the
> realm of theoretical discourse does bring certain theoretical
> advantages. For one thing, it provides clues to the functioning
> of our brains.

except that I'd want to put in various qualifications about what
sorts of things might be referred to by the phrase "one's
conscious experience".

Nevertheless, there are many phenomena concerning sensory
experience and ways in which it can change which provide clues as
to what's going on in the brain. Also when these experiences
change they can provide clues regarding the nature of damage to
the brain.

[Note added 22 Feb: as I remarked in a recent posting responding
to Stan Klein, even a robot with visual qualia might provide
clues as to functioning and malfunctioning of mechanisms deep in
its visual system if it could talk about its conscious
experience, e.g. reporting difficulties in binocular fusion.]

Some people who object to my functionalism seem to think that I
wish to define all functions in terms of impact on external
behaviour. I.e. they confuse functionalism with some form of
behaviourism.

In fact I think many functional roles of internal states and
processes can ONLY be defined in terms of their causal relations
to other INTERNAL states and processes. I.e. my form of
functionalism is inconsistent with all forms of behaviourism that
I know.

To characterise, in broad terms, this collection of internal
causal relations is to characterise an architecture. I.e. there
are non-physical architectures whose nature we need to
understand.

Although they are non-physical they are implemented in physical
systems. That's a kind of reduction, but does not involve
deducibility. [Note added: for the reasons explained in the notes
added in my previous message, because the webs of causal
relationships use an ontology not definable in terms of physical
concepts.]

I suspect there's nothing here that you disagree with.
[Note added: expect perhaps the point about non-deducibility.
See below.]

However, you want to go beyond this.

(a) You probably want to make additional statements about the
nature of these internal mental states. I suspect I may agree
with some of them and not with others. E.g. if you think zombies
can exist that have mental states and processes with all the
properties that I talk about yet lacking real seemings or
experiences then I find that incomprehensible: I have no idea
what it is that they are supposed to lack if they have everything
else.

(b) You want to make claims about the inadequacy of classical
physical systems to provide an implementation for these mental
states and processes.

(c) You believe that some sorts of mental phenomena are
intimately involved in the underlying dynamics of physical
reality (so that perhaps some of physical reality is implemented
partly in mental mechanisms???).

As regards (a) I have never understood what it is supposed to be
that zombies lack if they have internal states with all the
features and causal properties of desires, intentions, beliefs,
visual experiences, pains, pleasures, etc. that I talk about.

[Note added: I think many people who object to this sort of
viewpoint have not understood what sorts of functionally defined
mental states are at issue: they assume I am talking either about
observable behaviour, or about something like what happens in a
computer's CPU. They find it hard to conceive of an architecture
made of a rich network of high level unobservable dispositions,
which may or may not be implementable on a computer as far as we
can tell at present.]

As regards (b) I have never seen anything that looks like a valid
proof that there's an inadequacy in classical physics.

[Note: I've previously admitted that machines implemented using
only classical physics might turn out to be inadequate to support
human-like mentality -- but not for any reason that I've seen
proposed so far. Maybe there are other reasons, e.g. the need to
compress enormous information stores into very small spaces with
very low energy consumption, while preserving robustness?]

As regards (c) I don't know enough physics and mathematics to
understand the detailed arguments but I find it hard to
understand what sorts of mental reality or consciousness is
supposed to have existed in the physical world long before there
were any conscious animals, etc.

Aaron
PS
Assuming Henry does not object, I shall store this correspondence
starting with his original message, in:

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

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