From Aaron Sloman Sun Jan 19 03:07:15 GMT 2003
To: "PSYCHE Discussion Forum (Biological/Psychological emphasis)" 
Subject: Re: pre-requisites for discussing consciousness

Thanks for the comment.

mark jonathan horn wrote:

> I find Aaron's statements rather contradictory in spirit, in that I can see
> no significant connection between having "a fairly deep knowledge in a
> number of fields," and any profound importance of the software engineering
> or "virtual machine" paradigm as a primitive in relation to consciousness.

The lack of a connection is not a contradiction (unless you are
using 'contradictory' in a new way. Normally P and Q are said to
be contradictory if at most one of them can be true and at most
one of them can be false.)

I did not use the word 'primitive'. I stated that there are many
fields of research that are relevant to discussing questions
about consciousness. People who consider only a small subset of
those sources of knowledge can fall into conceptual,
methodological and empirical errors. (I had in mind particularly
unjustified claims about the need for reportability.)

I did not claim to give an *exhaustive* list of relevant fields of
knowledge.

In the context of discussions of reportability of conscious
states I put knowledge about virtual machines first partly
because so many people don't understand what they are, and partly
because most claims that conscious states must be reportable
derive from an out of date logical empiricist philosophy of
science. This philosophy of science (requiring detailed
testability of theories) has many counterexamples in the history
of science, and new counterexamples relevant to the study of mind
come from the developments I mentioned in science and
engineering which show how a virtual machine can contain complex
and sophisticated ongoing processes that are not reportable by
the system.

I gave some examples of different sorts of reasons for
non-reportability (insufficient bandwidth, no internal
self-monitoring, lack of an appropriate ontology for
self-categorisation, etc. etc.).

Those factors preventing reportability could apply to virtual
machines implemented in brains.

In the past such claims about virtual machines might have been
dismissed as metaphysical nonsense.

Likewise ideas about transmitting sounds and images through empty
space might have been dismissed as metaphysical nonsense before
we learnt about electromagnetic transmission.

None of this rules out alternative approaches to the study of
mind and brain in order to answer some questions: in fact I
listed several alternative approaches but did not claim
completeness.

> It seems that the strength of Aaron's general thesis on improving
> discussions on the concept of consciousness, is undermined by an overtly
> biased selection of the computational notion of "virtual machines" as
> somehow a more appropriate approach to consciousness than say, a quantum
> theoretic or relativistic model.

Should it have been less biased, so as not to be overly biased?
How much bias is OK?

If you can show examples of ways in which other sources of
information enhance discussions of consciousness I shall raise no
objections. Nothing I wrote was meant to be exclusive.

My comment on quantum mechanics was meant to be mildly ironic.
However, I have been warned in the past by expatriate Brits that
in the USA certain kinds of humor are not understood, but I had
forgotten. Sorry. (I was trying to show that almost everything
I was recommending had its dangers. E.g. I tried to indicate that
current theories about virtual machines were too simple.)

Quantum mechanics is very obviously relevant to the phenomena I
was talking about since minds and all other kinds of virtual
machines are implemented in physical systems, and (at present)
quantum mechanics is a fundamental part of physics.

The virtual machines built by software engineers make good use of
physical devices (e.g. transistors) whose functioning is
explained by quantum mechanics and this may be increasingly so in
future (See
    http://www.pbs.org/transistor/science/info/quantum.html
    http://www.sandia.gov/media/quantran.htm )

There are many aspects of brains and all other biological
mechanisms that depend on quantum mechanics.

Quantum computers may one day enable us to gain enormous
increases in speed, which could make a qualitative difference to
the implementation of human-like virtual machines that operate in
human time scales. That's an empirical question.

However there are many spurious (but published) arguments about
the relevance of quantum mechanics to consciousness, free-will
etc. that start from philosophically muddled analyses of the core
concepts. (E.g. I've written a lengthy critical review of some of
what Penrose wrote on this.)

Incidentally I did not say that virtual machines have to be
'computational'.

The notion of 'computation' is itself riddled with confusion and
ambiguity in our culture. E.g. for some people it includes only
digital computation, for others it includes analog computation,
and for some it is defined by a mathematical theory, whereas for
others it is defined by what various kinds of computers actually
do.

Causation is central to the last interpretation of computation
whereas the mathematical theory of computation is purely
structural and has nothing to do with causation. (e.g. the limit
theorems and complexity results apply to timeless structures:
formal systems).

In any case, I gave as examples types of virtual machines (e.g.
socio-economic virtual machines) that are not obviously
computational in any precisely defined sense.

It is probably going to turn out more useful to talk about
'information processing' virtual machines, where information
processing is something done by all biological organisms as well
as many artificial systems.

('Information' in the sense required here can no more be
explicitly defined than 'energy' can. Both are rich concepts
whose depth and complexity grow as our theories grow. E.g. Newton
did not understand the concept of 'energy' as we do now. It was
significantly extended by the discovery of chemical energy,
radiant energy, mass-energy etc. The concept is implicitly
defined by the theories using it, which can evolve over time.

Likewise, our notion of information (semantic information) is
implicitly defined by current theories about the kinds of
information content there are (including control information).
the kinds of relationships in which information can stand, the
kinds of things that can store or convey or manipulate
information and the kinds of things that can be done to
information (e.g. acquiring, storing, analysing, comparing,
describing (meta-information), combining, transforming,
interpreting, deriving, checking, transmitting, and above all
using it).

All organisms process information in this sense. But they vary
hugely in what information the process, the mechanims and forms
of representation that they employ and the uses to which they put
information.

Consciousness of anything (e.g. possession of qualia)
involves information in this sense.

Merely being conscious (not of anything) involves the capacity to
become conscious of various things, and therefore involves the
*potential* to have (and use) information.

Insofar as states, processes, and capabilities that constitute a
particular type of consciousness either are, or are closely
analogous to, states, processes and capabilities of
information-processing virtual machines, then since well
understood virtual machine states need not be reportable, there
is no reason why states of consciousness need be reportable. If
there is any reason why reportability is a requirement, I've not
heard it presented apart from the apparent assumption that state
descriptions need to be individually testable, which, as I've
said, is just bad (out of date) philosophy of science.

I tried to indicate examples of facts that could prevent
reportability. Several probably apply to humans.

> Given Aaron's view above that,
>
> >when we understand the space of possible virtual machines better we can
> >then refine and extend our ideas of consciousness to make them more
> >precise in different ways (relative to different architectures),
>
> I would suggest that this cannot be seen as any more compelling than the
> possibility that,
>
> "when we understand the space of possible conscious states better we can
> then refine and extend our ideas of virtual machines to make them more
> precise in different ways (relative to different architectures)."

Being 'compelling' is not relevant to the aims of science. What
we need in a theory is depth, precision, richness explanatory
power, and the ability to drive new research. (As suggested by
Lakatos in the reference I gave.)

Most good new theories were far from compelling when first
introduced (including quantum physics and relativity).

One problem with discussions of consciousness is that too many
'
things are found 'compelling' on the basis of uncritical
introspection. But people disagree on what they find compelling.

What is compelling is often false (e.g. the notion that we
understand simultaneity from our own experience of it). There are
many examples from the history of mathematics: e.g. it was felt
'compelling' that a curve could not fill a region, till one was
produced.E.g. see
    http://www.math.ohio-state.edu/~fiedorow/math655/Peano.html
There are many other examples. Some people probably find it
compelling that the rationals form a continuum, because there are
no 'gaps' between them.)

> If Aaron believes that computational construct of "virtual machines" is the
> best primitive architecture from which to launch more pregressive
> discussions on the biological and psychological implications of
> consciousness, why didn't he just state that up front?

I did not refer to primitive architectures. On the contrary the
architectures I did refer to are far from primitive.

Moreover, the notion that there is 'an' architecture
corresponding to the notion of a virtual machine fails to take
account of the huge variety of types of virtual machine
architecture.

We still barely understand that variety, even though we
understand a small subset very well because we know how to make
instances and put them to good use.

I don't see how I could have been more "up front" about the
importance of virtual machine architectures than I was. It was
point number 1 in my list, and was further discussed in an
extended note.

Maybe the phrase 'up front' has some meaning that is lost on me,
since we seem to using language differently.

Perhaps we are failing to communicate in other ways too?

Aaron