From Aaron Sun Jun 23 23:23:59 BST 1996
Newsgroups: sci.psychology.consciousness
Subject: Avoiding the noun "consciousness"

Many thanks to Stan Klein <klein@adage.Berkeley.EDU>
    Wed, 19 Jun 1996 23:03:22 PDT

and Bill Robinson (wsrob@iastate.edu)
    Fri, 21 Jun 1996 17:16:03 CDT

for responding to my remarks about the dangers of using the noun
"consciousness".

I'll respond to both in one message.

(I should have pointed out that what I wrote was largely unoriginal,
since related arguments were put forward as long ago as 1949 in
Ryle's book The Concept of Mind (widely misunderstood as a
behaviourist manifesto), and some of Wittgenstein's work, among
others. See also my comments on Kant and Frege below.)

First Stan's response.
[SK]
> ...
> I remain convinced however, that the neuroscientists have a decent
> chance of making some wonderful discoveries within our lifetime of
> what is special about those neural circuits that are connected with
> awareness vs. the more numerous circuits that remain unconnected.

"Awareness" is another of those dangerously deceptive nouns!

You can be aware
    - that it is getting darker,
    - that you are making a fool of yourself,
    - that someone is following you,
    - that you are getting tired,
    - that you have not expressed yourself clearly,
    - that the person you were following is no longer in your field
      of view,
    - that the chess move you have been thinking about is dangerous,
    - that you are infatuated with someone you have met,
    - that your infatuation is distracting you from
      important tasks,
    - that the current political climate is becoming dangerously
      extremist and intolerant,
    - that there must be a flaw in the proof you have thought up,
and many more.

Do you have any reason at all to believe that there are some special
circuits that are connected with ALL these things? Or are you
thinking only of a special subclass? (Compare Bill).

If the latter, then it may be better to define the subclass you are
studying with some care, instead of using words like "awareness" and
"consciousness" whose application is far more general.

(One of the problems in studying the brain is that many of the
important circuits are not direct links along a single nerve fibre,
but indirect. However, as soon as you allow indirect connections,
then everything is connected with everything else, in which case
*everything* is connected with awareness in some circumstances!)

[SK]
> I would love to be able to convince Aaron that the search for the
> NCC is an exciting area. But I know full well that what is exciting
> for one person is pretty boring to another.

If I were merely bored with the discussions of consciousness I could
(and would) ignore them, as I ignore lots of other research.

However a great deal of what is written is not so much boring as
conceptually confused and pernicious because it gives people who
should know better the impression that there's some holy grail they
should be seeking.

[SK]
> I think Aaron is raising the fundamental issue of whether there are
> neural correlates of consciousness that can be discovered by the
> neuroscientists.

No. My problem is more basic than that.

I am suggesting that many (not all) people who talk about the search
for correlates are deluding themselves in thinking that they have
identified a research task, because they think they know what they
are referring to on one side of the correlation, when they don't.

They are implicitly presupposing uniqueness of reference of nouns
like "consciousness", "awareness", "experience" (and sometimes
others, e.g. "subjectivity").

[SK]
> The answer isn't in yet, but there is recent
> exciting work on the neural correlates of attention and neural
> correlates of binocular rivalry that make one optimistic.

Optimistic about what?

Of course the search for neural mechanisms underlying all sorts of
mental phenomena is important and should be pursued diligently.

This includes the study of mechanisms of binocular fusion and
binocular rivalry (and the fascinating discovery reported in a
recent BBC TV programme that colour-blindness can be "cured" by
providing a contact lens for one eye that's tinted with a colour
specially tuned to the subject's colour blindness: which shows that
colour perception can be binocular.)

But do you have any reason to think there's going to be something
common to all these:
    - the mechanisms of binocular rivalry,
    - the mechanisms of detection that your foot is cold,
    - the mechanisms underlying a student's first coming to
      understand Euclid's proof that no rational number can be the
      square root of 2,
    - the mechanisms underlying the appreciation of a Shakespeare
      sonnet,
    - the mechanisms involved in feeling hungry or humiliated or
      experiencing an orgasm or remembering an event in your youth,
        etc. etc. ?

Of course, all these mechanisms are in the brain.

But so are the mechanisms I referred to underlying processes of
which we are NOT aware, e.g. mechanisms in speech understanding and
generation, learning by abstraction, blindsight, control of finger
movements in touch typing, etc. Do you have any reason to suppose
that there is some fundamental (non-relational) difference between
the two classes of mechanisms?

(A relational difference is concerned with functional relationships
to the rest of the system, like being an employee, or being married,
or being entitled to vote, or owning a house.)

Bill Robinson raises a different set of objections, from a
philosopher's viewpoint, to my modest proposal:

[BR]
> Aaron Sloman suggests (psyche-d list, 6/19/96) that we eschew use of
> the noun "consciouness" for five or ten years. But this is an
> unscientific suggestion that would be pointless.
>
> Pointless -- because the key issue can be raised without the noun.

Yes of course!

I did not think for a moment that people could *really* do without the
noun, or that anyone would take my modest proposal seriously. I've even
been known to use the noun myself!

I was trying to draw people's attention to the dangers of unjustified
ontological assumptions which could be made unwittingly because of the
delusions so easily induced by linguistic forms, especially noun
forms.

This is an old idea, as philosophers will know: one of the earliest
examples being Kant's attack on the notion that "exist" is a
predicate, e.g. in "God doesn't exist". Another is Frege's critique
of the notion that number words name properties of objects: as might
be suggested by the similarities between such pairs as "What colour
cats are in this room?" -- Answer "black". "How many cats are in
this room?" -- answer "fifteen". Frege (unlike Mill) suggested that
one is a question about properties of the cats and the other a
question about a second-order property of the set that contains the
cats, or of the predicate "is a cat in the room".

[BR]
> ...In
> fact, Sloman provides an excellent way of raising it, with his list of
> things of which we are NOT conscious (e.g., several aspects of the
> processing necessary to support speech). Surely, we can sensibly ask
> what neural properties distinguish this class of occurrences from
> occurrences in which there are conscious experiences (note the
> adjectival form).

Yes, but what reason is there to believe that there are any
*properties* as opposed to relations that distinguish the two
classes?

My whole point is that many people (especially people with no
experience of designing complex working systems) unthinkingly assume
there *must* be some common neural *properties* whereas I was trying
to open up their minds to the thought that the only common things
might be high order *relationships* within a larger system, and some
aspects of the relationships would be common to mechanisms
underlying both accessible and inaccessible mental states and
processes.

(The differences between being aware of a pain in your toe and
grasping a mathematical proof or appreciating a poem are
sufficiently great that I don't expect there to be a single simple
relationship either. It's more likely to be a cluster of overlapping
cases.)

[BR]
> ..But that is a way of asking the same question as is
> intended by asking for the neural correlates of consciousness.

And just as misguided, if it starts from the unquestioned assumption
that there must be, or are likely to be, common neural *properties*.

Of course, there might be: it's an empirical question. But I have
not yet seen the slightest shred of evidence to support that
hypotheses, given the diversity of structures and processes in the
brain that seem to be involved in various kinds of mental events
some of which we are aware of some not. And as a
philosopher-engineer who believes that the brain is an exquisitely
designed information-based control system, I can't see how any
neural *property* could have the right explanatory powers.

Neither, as far as I can tell, can anyone else.

[BR]
> Sloman suggests that the neural correlates of conscious events may
> not differ in any interesting way from the neural correlates of
> nonconscious events.

NB: saying their intrinsic properties may not differ is consistent
with saying that there might be interesting differences in their
neural relationships, as I suggested in my message.

[BR]
> How this matter stands is a scientific question; it
> is thus unscientific to suggest that we adopt *in advance* of
> investigation a linguistic policy that's designed to obscure one of
> the possible outcomes (indeed, the one that seems most natural and
> likely to some).

I agree that it is an empirical question. I am not a brain scientist
but everything I have read and heard about the brain, together with
everything I know about the problems of designing sophisticated
autonomous information-based control systems makes it look to me as
if this outcome is SO unlikely that it would be far better for
people to search for other possible outcomes.

(Many people working in the field don't seem to grasp the conceptual
framework required for some of the interesting alternatives. Or if
they do, they don't show it. There are notable exceptions. E.g. I
think Dennett understands the point but causes confusion by saying
that he is explaining consciousness, as if there were some *thing*
to be explained! And he does not do justice to the design
requirements for a system that can explain why people want to talk
about qualia, or why intelligent robots will do so.)

I would not object if the scientists claiming to investigate the
neural correlates of consciousness (or awareness) actually started
by formulating the obvious main alternatives for what the correlates
might be (e.g. common property vs analogous functional
relationships) and then carefully distinguished the types of
evidence that might discriminate them.

I suspect that for some the problem is just that people whose
philosophical training has been limited are not aware of the
unjustified inferences they make (unconsciously) from linguistic
forms. For others, it's limited experience of solutions to very
complex design problems.

[BR]
> Sloman has a persuasive argument when he points out, rightly in my
> view, that we should not expect to find any interesting neural
> properties that correlate with knowing a fact or having a skill, or
> interesting electronic properties that correlate with being a
> compiler detecting a syntax error, etc. The latter members of these
> pairings are functionally specified properties that we can see will
> have many realizations, and there is no reason to expect the
> different realizations to have much in common at the neural (or,
> electronic) level.
>
> If we knew that the properties of conscious experiences, like
> painful, itchy, green, oregano-odorous, etc. were functionally
> specified properties, we could be sure there is an analogy here that
> might dissuade us from looking for neural correlates of conscious
> experiences.

Bill is implicitly doing what I was recommending, namely
distinguishing some subclasses of "consciousness" from others and
suggesting that there might be different neural explanations. Thanks
Bill.

Not everyone is so sensitive to these distinctions.

[BR]
> As it is, however, it is contentious to hold that
> painful, itchy, etc. are functionally specified properties;

I agree that it is contentious.

I am trying to elaborate a conceptual framework in which I can
present the case while avoiding all the more common objections. But
there's a long way to go.

I've made a start in a draft paper called "Actual possibilities"
which attempts to present a view of objects of all kinds as
"possibility-transducers", with properties that summarise
possibility transduction capabilities. (E.g. electrical resistance
of a conductor summarises relationships between possible voltages
and possible currents.) Complex systems link together objects in
such a way as to link and constrain the sets of possibilities that
can "flow" through the system, i.e. complex systems are possibility
transducers constructed from simpler ones. Some of the outputs of
sub-systems can change the possibility-transduction properties of
other sub-systems (or themselves). These descriptions can be applied
at different levels of abstraction (physical, neural,
information-processing, economic, social, etc.)

[There are some problems about what the contents of these
possibilities are: e.g. is it possible for objects to have
infinitely precise position, mass, resistance, electric current,
etc.? If not, this leads to a view of classical physics in which the
laws and equations are fully deterministic but the states (boundary
conditions) may be inherently indeterminate (i.e. not infinitely
precise -- which does not mean that they are discrete). Chaotic
systems (i.e. non-linear feedback) can amplify the indeterminacy
arbitrarily, but global structure can constrain outcomes (e.g. the
discretisation forced by the geometry of gambling devices). Do we
know what sorts of high level functionality can and cannot be
implemented in such systems?]

In more complex, self-modifying, networks of constrained
possibility-transducers, important classes of functional
relationships can emerge that have no place in simple electronic or
neural circuits (e.g. recent connectionist models).

In this framework I suspect things like itches, pains, pleasant
tingles, colour experiences, and other sorts of sensory qualia
(Bill's second class), can probably eventually be accommodated in
terms of temporary high level control states with particular sorts
of causal powers, closely coupled with and constrained by low level
sensors and monitors of various kinds.

The qualitative content of such a state (green vs purple) will be
nothing more than a relationship between the actual state and the
class of functional possibilities from which it is "selected". (I
know some people find that highly implausible, but I conjecture that
robots built on these principles will also find it implausible that
that's how they work, because of limited access to the true nature
of their own states.)

However, noticing a flaw in a mathematical proof or suddenly
remembering that you forgot to buy the bread will be very different,
both from each other and from the sensory states. They are instances
of quite different ranges of possibilities.

E.g. the content of the state of being aware of a logical flaw in a
proof is quite unlike the qualitative content of experiencing a
particular colour, though they may have other things in common, e.g.
like starting at a particular time, grabbing attention, producing
pleasure, displeasure or curiosity etc. ....

I am still struggling to put these ideas (and their implications for
functionalist theories of mind) together in a coherent form that is
easily understood by others. So far without much success. The draft
paper is too long to post but can be found at

    http://www.cs.bham.ac.uk/~axs/misc/real.possibility.html

I need help from modal logicians, among others!

There are many other issues that need to be addressed, not raised in
that paper, e.g. objections by Putnam and others to naive
functionalism (i.e. relational theories that depend only in
*internal* functional relations). This requires some aspects of the
(semantic) content of functional states to be defined in terms of
relations to the environment. (This defeats twin-earth arguments,
etc.)

[BR]
> and so
> to accept that Sloman's cases constitute an instructive analogy
> would be to take a substantive position on the key issue in dispute.
> Again, this is not something we should do in advance of
> investigation.

OK. But the opposite: namely the implicit assumption that the
analogy is misguided, as expressed by the use of phrases like "the
neural correlates of consciousness" is just as bad.

It's time for a division of labour. Let those who so desire search
for NCCs. Let others explore explanatory architectures and
relationships within such architectures without assuming there will
be any common distinguishing explanation for the phenomena referred
to in ordinary language by the noun "consciousness".

I'll put my money on the latter (and my research efforts).

Cheers
Aaron
===