THIS FILE IS
http://www.cs.bham.ac.uk/~axs/misc/classicalphysics-and-mind.txt

It contains a number of messages composed in answer to a long message
by Henry Stapp. Henry was outlining some of his reasons for thinking
that classical physics is an inadequate basis for explaining
mental states and processes, and objecting to some previous
arguments by Pat Hayes and myself.

In this collection of responses to Henry I attempt to show why I think
his characterisation of classical physics is incorrect and why his
arguments to show that mechanisms using CP cannot provided an adequate
basis for mentality are incorrect.

One crucial point is that in his message, as in various publications, he
characterises CP as concerned with purely geometrical facts about
trajectories of particles. I argue that that leaves out causality.
Because of the role of causation of various kinds, CP it is not purely
geometric. Causal relations include a rather intricate web or relations
which can hold at many levels of abstraction in a hiearchy (or partially
ordered collection) of ontologies implementing other ontologies.
(Implementation and supervenience are closely related.)

Some of the other points include the differences between logical
implication and causal necessitation, and the difficulty of attaching an
adequate semantics to modal operators expression notions of causality.
(Analysing our ordinary notion of causation is probably the hardest
unsolved problem in philosophy.)

There are related point in an older, incomplete, working paper on
supervience and implementation,
    http://www.cs.bham.ac.uk/~axs/misc/supervenience

There are lots more relevant papers in the Cognition and Affect project
directory:

    http://www.cs.bham.ac.uk/research/cogaff/


=======================================================================


From Aaron Sloman Sun Nov 14 10:39:37 GMT 1999
To: Henry
Subject: Re: Reply to Henry's reply (different from Pat's) [Part 1]

Hello Henry,

Thanks for your considered response to what Pat and I wrote. Despite
the large amount of agreement between Pat and me, we differ on
details, which I think are important though maybe some would say
they are merely differences of emphasis. So my comments are
different from Pat's.

I think I can now diagnose more clearly why I reject the description
of classical physics as essentially concerned with geometrical
configurations. That description leaves out the crucial importance
of various kinds of causation. I expand on that below.

But first some background on the question you want to discuss and
the questions you don't want to consider which I think are more
important, as they lead to deeper and more accurate answers to the
high level questions from which you start.

[HPS]
> I affirm that the question at issue is a logical one. It is:
>
> How can the notion of ``entail'' be understood in a way that allows one
> to claim, validly and rationally, that, within the framework of classical
> physical theory (CPT),
>
>
> 1) Micro-physical facts can entail such macro-physical facts as
>    hurricane facts, and airplane facts, and locomotive facts,
>
> but
>
> 2) Micro-physical facts cannot entail the existence of
>    felt qualities such as painfulness and
>    subjective (phenomenal) redness.

In order to clarify what's wrong with some of your subsequent
comments I need the following three reminders (a) to (c).

Indepenently of whether CPT or QM is true, the following are true:

(a) Micro-physical facts can entail many macro-physical facts though
possibly not all. Newton showed by mathematical reasoning how to
"integrate" the effects of forces acting on large numbers of rigidly
connected micro particles forming a macro-physical object. But it's
not obvious that all features of all macro objects can be dealt with
that way.

E.g. if there are certain macro-physical facts that use concepts
that are NOT definable in terms of the micro-physical concepts, then
there will be facts expressed using those concepts that cannot be
entailed by the micro-physical facts (and laws) alone.

Whether "hurricane", "leaf", "ocean", "oak", "elephant" are such
irreducible concepts is a matter of dispute. Pat and I would say
they probably all are irreducible, but I don't think Henry (or Stan)
agrees that they are. So, discussing these examples simply confuses
the issue because we are not clear, and certainly not all agreed, as
to what they are examples of.

Point (b) introduces clearer examples.

(b) There are many types of NON-physical facts which cannot be
logically entailed by micro-physical facts, including

    (i) facts about events and causal interactions within information
        processing virtual machines,
   (ii) facts about social, economic, legal, political events,
  (iii) facts about mental states and processes.

These involve (are implemented in, are supervenient on)
macro-physical mechanisms but should not be confused with
macro-physical facts which use concepts definable entirely in terms
of micro-physical ones.

Lets distinguish "pure macro-physical facts" and "hybrid
macro-physical facts". The pure ones are no problem. We all agree
that they are logically derivable from micro-physical facts. So
let's not talk about them.

Everyone would say that hurricanes, leaves, oceans, oak trees, and
elephants are all physical objects, whatever else they are, and many
events involving them are physical events (e.g. elephants walking or
eating or falling into elephant traps). Whether facts about them are
pure or hybrid is unclear. So lets not talk about them either, lest
we get bogged down over examples instead of the central issues.

By contrast, the events of types (i) to (iii) are NOT physical
events, not even macro-physical events: multiplication of numbers,
correction of a spelling error, parsing an html expression,
capturing a pawn, getting into debt, the end of an economic
recession are not physical events.

Non-physical events can be implemented in (or in philosopher's
jargon, "supervenient on") physical events without being physical
events.

(Pat sometimes writes as if they are physical -- e.g. saying that
the non-physical and the physical entities are identical, i.e. the
same things. He and I then part company: I don't find that concept
of "identity" useful -- but it is probably a disagreement of
terminology rather than of substance, since it's a question of how
to delimit the use the notion of "identity".)

(c) There are non-physical facts (like those referred to in (b)
whose existence is not logically entailed by the existence of any
physical facts, but whose existence is CAUSED by physical facts. (To
be more precise they involve events and processes whose occurrence
is caused by physical events and processes.)

(This notion of causation is used by all of us all the time, but it
is VERY hard to analyse. Hume tried and failed, among others. It is
intimately connected with truth of counterfactual conditionals.
Analysing the concept "cause" is the hardest unsolved problem in
philosophy.)

Henry says he is not asking questions about such non-logical
relations.

However if the relationship between mind and body is causal and not
logical, then people who do not pay close attention to causal
connections which are not logical connections will NEVER understand
the relationship between mind and matter (or between different
aspects of mind, or between social and psychological phenomena, or
between computational virtual machines and physical machines, etc.
etc.) Consequently they will remain puzzled about the relation and
then spend many years search in the wrong place for the
relationship.

Thus ignoring the questions I pose is risky: you may be missing
some truth and looking on the wrong place.

My claim (c) is superficially very close to what Dave says about the
relation between facts about consciousness and physical facts: i.e.
he thinks there is some basic, inexplicable, incomprehensible law of
nature as a result of which there can be non-logical causal links of
a unique kind between physical and mental phenomena.

He and I agree on almost everything except the uniqueness (and
perhaps the mystery): I claim that many different kinds of
non-physical phenomena are implemented in physical phenomena and are
thus causally related to them.

So the physics/experience link is just one example among many, not a
special and mysterious link.

It is also not an epiphenomenal relationship because causation goes
both ways, as explained below.

Morever, even when the causal links are not logical we can, in many
cases, understand, after a fashion, how they work: they are not
mysterious. Engineers use such knowledge all the time.

How we understand the links between levels of abstraction in such
engineering designs, and why they involve a kind of necessity which
is not logical necessity are important issues on which I think we
are not yet clear, but that's something I'll not discuss today.

Moreover, everything I have said so far is completely independent of
whether Classical or Quantum physics is true. Only the fine details
differ between CP and QP.

[HPS]
> I shall answer this question paying close attention to the very reasonable
> points raised by Pat and Aaron during the six weeks of discussion.

You are too kind....

> First there is the question of ``de re'' versus ``de dicto''
> raised by Aaron's Sept 26 message: He asks whether I am speaking about a
> fixing or determining of `actual facts of nature' by other `other actual
> facts of nature', or, instead, about logical implication within a theoretical
> framework.
>
> The answer is definitely the latter.

That's certainly what you *want* to talk about. That's clear.

My claim is that by restricting yourself to that viewpoint, by ignoring
the ``de re'' causal relationships, you'll blind yourself to the correct
answers, and therefore be attracted to spurious answers to your high
level questions about the relationship between mind and matter

[HPS]
> We know that the world itself is
> definitely not describable in terms of the concepts of CPT: that framework
> is an invention of man that is contradicted by the facts. Hence the
> entire logical structure under consideration must be understood as existing
> within a certain theoretical framework that is known to be incompatible
> with nature herself.

However, that is independent of the question whether CPT if true,
could causally entail (``de re'') facts of a type which are not
logically implied (``de dicto'').

Consider this: I think you believe (as Stan, Pat and I and others
believe) that if CPT were true it would be possible to construct
computers which do most or all of the things existing computers do
(though maybe such computers could never be as small, fast, cheap,
reliable, and low in energy consumption).

If you agree with that, then you must agree with me that physical
events in such a computer using only classical mechanisms, could, if
it were suitably programmed, cause the existence of certain
NON-physical events, such as addition of two numbers, or capture of
a queen in a chess game, or correcting spelling in an essay.

I.e. the causation of such events in computational virtual machines
implemented in physical machines does not depend on the truth of QM.

Moreover, the relationships between the physical events and the
virtual machine events are not logical. (I sometimes think Stan
thinks they are!) They cannot be logical if the description of the
events essentially uses concepts not definable in terms of those of
physics (e.g. "addition", "queen", "spelling".)

Morever the sense in which such a relationship between events in
different ontologies (the relation of supervenience, or
implementation) could hold in a CPT world is exactly the same as the
sense in which such relationships do hold in a QM world.

I.e. the following is true:

(d) The mere fact that CPT is false and QM is true is irrelevant to
the question whether whether certain things could be causally
necessitated by classical mechanisms.

Moreover insofar as mental events are causally necessitated by
quantum mechanisms in brains or computers, this need not depend on
any of the special features of QM, e.g. non-locality, indeterminacy.

The causal necessitation depends only on features which quantum
mechanisms share with classical mechanisms, e.g. the features that
would allow computers to be built in both. But what are those
features? They are not purely geometrical features.

[I don't think Pat likes this talk of causal necessitation: He
prefers all necessitation to be logical. Curiously, I think that in
this he is a bit like Henry???

Things Stan has written about his belief that various brain
phenomena use only classical mechanisms seem to agree with (d).]

[HPS]
> The issue is about logical implication within this specified theoretical
> framework: it is stricly ``de dicto''.

That may be all you think you are interested in. But there are other
relations you *ought* to be interested in, IF you really wish to
understand the relation between mind and matter (or computation and
matter, or economics and matter, etc.).

It is because you ignore these other possibilities (causal
connections) that you think ONLY QM can support and explain mental
phenomena.

So have a go and look at the questions you think are of no interest!

If you don't want to then maybe we have nothing more to discuss. But
I think your mind is more open than that, so....

My next message is about the need to avoid wrong models of causation
and why classical physics is not purely geometrical, i.e. because
it, like QM, includes causation, which has profound implications,
not yet fully explored or analysed.

Aaron


From Aaron Sloman Sun Nov 14 11:29:54 GMT 1999
To: henry
Subject: Re: Reply to Henry's reply  [Part 2] (multi-level causation)

We shortly come to the reasons why classical physics is not purely
geometrical, and why that's so important. But first a point about
completeness of causation.

[HPS]
> When I speak of the `description of the microphysical facts', I am doing
> so from the perspective of a theoretical physicist who is using CPT:
> the complete description of the `microphysical facts', within the
> CPT framework of thought, is supposed to be a complete description of the
                                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> trajectory in spacetime of every particle in the universe, and also the
  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> description of the value of every physical field at every spacetime point.
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

I return to the difference between the underlined bits later. The first
refers to geometrical facts, the second to causal powers.

Facts about these causal powers are not geometrical facts.

> This (imagined) micro-physical structure is one key component of the
> theoretical framework that constitutes CPT.

So a hypothetical computer built on purely classical principles
could have such a description, in terms of its geometrical
configuration, the various fields, the physical states of the
micro-physical components, etc., and in one sense that would be a
``complete'' description.

In particular, the description would leave out no facts involving
the micro-physical ontology, and all events at that level (in
that ontology) could be completely causally explained at that level.

(For now I am ignoring the indeterminacy which would follow from
lack of infinite precision plus chaos, even in Classical mechanics.)

In another sense the description would be incomplete, since it would
leave out facts using concepts which could not be defined in terms
of physical concepts, e.g. the fact that it is running a prolog
program, or playing chess, or adding numbers. These are not physical
facts.

There would also be both pure and hybrid macro-physical facts
concerning the whole computer, but we can ignore those for now.

It is a very important feature of the world that in general what is
going on in any particular region of the world can be described at
many levels of abstraction, and causal relations can hold both
within and between all those levels of abstraction, in all
directions.

Many years ago, Elizabeth Anscombe, in her little book "Intention",
discussed in detail an event in which some matter moves up and down,
(partially) implementing an arm moving up and down, (partially)
implementing a person operating a pump, (partially) implementing a
process of refilling a water tank at the top of a building,
(partially) implementing a process of transferring poison to the
water tank, (partially) implementing a process of poisoning the
inhabitants, (partially) implementing an act of revenge, etc. etc.

I've used my own way of describing the example, not hers.

Such examples point to the fact that the "topology of causation" is
very rich and complex, partly because the ontology within which
causal connections hold is very rich and complex, and multi-faceted.

Some people think that causal completeness at the lower levels
implies that at best the upper levels must be epiphenomena: lacking
in causal powers, or at least lacking the ability to influence the
lower levels.

That's a mistake: despite causal completeness at the micro-physical
level, some of the more abstract non-physical events can be causes
of the micro-physical events.

I.e. a queen being threatened could cause a defensive move to be
made, and since the chess virtual machine is implemented in physical
mechanisms, when changes occur in the virtual machine (e.g. a queen
is threatened, numbers are multiplied, spelling is corrected), they
cause physical events. This is an instance of a very old and for
humans essential way of thinking about the world, e.g. when we say
poverty can cause crimes, including crimes which involve movement of
physical objects, or the desire for revenge caused poison to be
added to a water supply.

In our ordinary life we take such top down causal influences for
granted: indeed we could not live as we do without doing so. But in
our philosophical armchairs we impose an incorrect theory of
causation and then want to reject such causation (or explain it by
irrelevant appeals to causal gaps in micro physics, or by adding
mentality to micro-physics, in Henry's case.)

Conjecture: the existence of causal connections depends ONLY on
the truth of certain sorts of conditional statements, including
certain sorts of counterfactual conditionals. Such truths can
underpin causation can going top down, horizontally, middle up, etc.

If the queen had not been threatened the computer would not have
chosen that defensive move, and those transistors would not have
flipped their state. The queen was threatened and the computer
therefore chose the defensive move. So the threat to the queen
caused physical changes.

This does NOT imply that the physical machine had any causal gaps.
In fact, if it had causal gaps the system could not work as it does.
So micro-physical events that are completely causally determined by
micro-physical events can also have non-physical causes.

When I say "the truth of certain sorts of counterfactual
conditionals" this needs to be expanded in such a way as to prevent
events in a mirror image causing events outside the mirror. I don't
yet have a complete analysis of all this.

Anyhow, given the above conjecture and a few other simpler
assumptions, it follows that causation is always highly
over-determined. There are no unique causes.

That's because different ontologies co-exist in intimate
connections, as illustrated by Anscombe's multiple descriptions of
what's happening in a particular location at a particular time.

Most people wrongly believe that *completeness* of causation implies
*uniqueness*. This can drive them to epiphenomenalism: which allows
physical events to cause non-physical events but not vice versa.

I think this mistake is made because people wrongly think of
causation as a kind of ``fluid'' that flows round mechanisms subject
to some sort of conservation law. (Like billiard balls transmitting
momentum when they collide.)

However causation is a far more subtle, complex and multi-faceted
concept, deeply connected with the idea of "what would happen if".
That is why circular causation is possible. If that is not agreed,
or not understood, then the rest of what I say will fall on deaf
ears.

Given all that, Henry's search for *logical* connections between
physical and mental phenomena can be seen to be misplaced as a route
to a unified theory of the world.

We cannot make do with logical (or, more generally, mathematical)
connections in order to explain the coexistence and deep mutual
interaction of mind and matter, or matter and matter for that
matter. We need a theory of causation. Among other things, we need
to understand what I've tentatively, and perhaps misleadingly,
called the "topology of causation".

There are some models of causation which have the topology of a
branching network of possible worlds with a (possibly non-unique)
successor relation. Treating each possible state of the universe
as a node in such a simple graph, does not do justice to the need to
overlay multiple ontologies in any particular world.

Different coexisting ontologies support different kinds of branches.
E.g. there are far fewer moves possible in a chess position than
possible state changes in the computer that implements the chess
game.

Some ontologies may not allow discrete states, e.g. the Newtonion
ontology of planetary motion and a faraday/maxwellian ontology of
electromagnetic phenomena. Yet discrete ontologies can be
implemented in continuous ones in classical physical mechanisms.

I keep saying that analysing causation is the hardest unsolved
philosophical problem. That includes analysing what it is to
understand a causal connection.

E.g. when you open up an old clock and see how it works, or examine
a computer program and understand why it goes wrong. Sometimes there
can be a lot of mathematics involved. But the mathematics has to be
combined with an understanding of something else: causal links.

I am not sure we can ever get a fully detailed and complete
understanding of ALL types of causal relationships, because of the
human impossibility of grasping, or specifying, all relevant
boundary conditions, not even at the physical level.

I've discussed that before, but I need to find a way to spell it out
more clearly.

However, this limitation doesn't stop us having a rich and deep and
powerful understanding of some of the causal connections that link
ontologies: the sort of understanding that enables us to build and
sometimes debug computers, even if some of their behaviour defeats
us.

I'll return later to the way in which classical physics involves
causation, and how that takes us beyond geometry.

But first my next message is a digression about the need to separate
out questions about the nature of classical physics from questions
about the experience of classical physical phenomena. Classical
physics, like Euclidean geometry, goes way beyond experience.
Concept Empiricism is a false philosophical theory.

Aaron


From Aaron Sloman Mon Nov 15 03:17:50 GMT 1999
To: Henry
Subject: Re: Reply to Henry's reply  [Part 3] (CP, causation and experience)

[HPS]
> Newton, in ``Principia'', tied the micro-physical description to macroscopic
> phenomena, such as the tides, which properly trained people can observe.
>
> Classical physical theory is used by scientists to form useful and
> testable predictions about all sorts of observable phenomena, from
> meteorology, to aerodynamics, to steam engines. So how is this
> theoretical construct, which is expressed in terms of microscopic
> properties invisible to human beings, connected to our actual and
> possible experiences about the world.
           ^^^^^^^^^^^^

When the macro phenomena are *definable* in terms of agglomerations
of the micro phenomena there's no problem explaining how they are
connected to the micro phenomena. E.g. motion of planets and
collisions between large rigid elastic objects might be such macro
phenomena.

The micro/macro connections are then logical (or more generally
mathematical, including for instance the kind of mathematics used in
integral calculus or solving differential equations: which goes
beyond what's normally called "logic").

How EXPERIENCES of those macro phenomena occur is a totally
different question -- to be answered by producing a design for a
fully functional mind in which experiences can occur.

It's important not to confuse the two questions: (1) how are such
macro phenomena explained and (2) how are experiences of the macro
phenomena explained.

Certain philosophers (e.g. Bishop Berkeley, and sometimes Henry)
think nothing can exist without being experienced. But that flies in
the face of what most people rightly think (most of the time).

There are all sorts of things going on all around the universe that
are not experienced by anyone or anything. (If Henry really thinks
that is logically impossible, then we probably can't continue this
discussion without a huge digression: it seems to be his view when
he writes about QM, but I am not sure whether it's offered only as a
scientific conjecture or something stronger.)

I believe we can one day explain how macro-physical phenomena are
experienced, by designing the architecture for a mind which is
capable of having those and other experiences, but we still have a
few more years of work to do before that task is complete!

Anyhow, the main point for now is that it is a mistake to confuse
the explanation of the occurrence of macro-phenomenon X with the
explanation of the experience of macro phenomenon X. (Each can occur
without the other: prehistoric events, and hallucinations.)

[HPS]
> Classical physical theory is erected on idealizations of certain aspects
> of the complex and many-faceted image of the physical world that arises
> in our stream of consciousness.

This apparently expresses an empiricist philosophy of science: "all
concepts derive from experience".

It's not the only philosophy of science. Many philosophers of
science (e.g. Popperians) nowadays accept that science uses
ontologies that go beyond experience. This would be true of the
forces, fields, momentum, energy, etc. of classical physics, as well
as concepts of sub-atomic particles, electrical resistance, magnetic
properties, etc.

Moreover, it's not always noticed that the purely geometrical kind
of "idealization" referred to also goes beyond anything experienced:
such idealisations include conceptions of infinitely small points,
infinitely thin or straight or long lines, etc. Nobody has ever
experienced anything shrinking or stretching to the limit. Nobody
can experience the difference between a perfectly straight line and
a short segment of the circumference of a huge circle.

How humans perform such mathematical idealisations, whether any
other animals can, why some others cannot, whether very young
children can, whether the ability is innate or bootstrapped, how it
evolved, etc. are all fascinating unanswered questions in
psychology, brain science, biology, etc.

Hence, if the geometry of the space in which we perceive our
environment is euclidean geometry, it is not an experienced
geometry. What we experience is only certain aspects of that
geometry, but not other aspects, e.g. we cannot experience Euclidean
points and lines, inifinite volumes, *perfect* straightness,
*perfect* spheres and circles, etc.

These were issues that exercised Kant, who concluded that our
understanding of space could not come from experience since it was
needed in order to have experiences in the first place.

He concluded that geometrical concepts were innate. However, he
didn't pay any attention to the fact that different aspects of
geometrical understanding are innate in different animals.

Presumably what's innate uses information discovered by evolution
and encoded in genes.

In some organisms, it seems that part of the information is not in
the genes, but is developed and encoded by indidivuals through self
organising perceptual mechanisms, after birth or hatching.
(Altricial species.)

Either way, geometrical information required by biological organisms
was being processed (in ways not yet understood) long before there
was any consciousness of geometry. So it's wrong to think of
geometry as something that in any way *depends* on experiences, even
if some of what *we* know about it arises out of our experience.

Likewise I can see fragility, rigidity, stability, in things in the
environment (Gibsonian affordances) but the things I see are not
experiences in the way I think Henry conceives of experience. They
are all things that can exist without being experienced. Eventually
we discover them and learn to recognize them in subtle ways, acting
as individual scientists, and even later produce mathematical
theories about them.

[HPS]
> Descartes,  who was not only the founder
> of modern philosophy, but the also the creator of analytic geometry,
> picked out the GEOMETRIC ASPECT, namely the features characterized by
> extension in space and time, as the basis of a conception of
> `that aspect of nature that persists when no one is experiencing it'.

Good for him.

> According to this conjecture of Descartes, there is an aspect of nature
> that exists outside of anyone's experience, and that aspect has a
> microscopic spacetime structure: it can be described in terms of properties
                                                                   ^^^^^^^^^^
> assigned to spacetime points.
  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

But not only geometric properties. There's also electrical
resistance, or charge, or voltage, or degree of magnetisation? These
are all part of the classical physics I was taught as a youngster.

All of these involve causal powers. Causation is not a geometrical
relation, though it can depend on geometrical relations in some
cases. (Inverse square laws, etc.)

Within a purely classical world (if it existed) there would be causal
connections between events, in addition to spatial, temporal,
geometrical relations.

Once we have causation in addition to geometry we can build machines
that DO things. Geometry (in its original sense: Euclidean geometry)
says nothing about ACTION or any form of CAUSATION. It is only
concerned with STRUCTURE.

Some machines based on purely classical physical principles can do
information processing. If suitably designed, they can implement all
sorts of non-physical events and processes, such as addition,
multiplication, searching for information, logical inference, etc.

Geometry is not enough for this, not even in classical physics. The
previously underlined quotation from Henry (underlined in mhy
part 2 message) using the phrase "the value of every physical field
at every spacetime point" implicitly acknowledged this.

Fields involve causation, not just geometrical entities and
relations.

[HPS]
> This conceptualization is based on idealizations that were already the
                            ^^^^^^^^
> basis of Euclidean geometry,

But the classical conception of physics is not based ONLY on those
idealizations. Euclidan geometry says nothing about causation. It is
not even mentioned in any of the axioms of geometry as far as I
recall.

Causation comes into classical physics through notions of force
(gravitational, electrical, magnetic, elastic, ....), energy
(kinetic, potential, chemical, thermal, optical, ....) momentum,
stability, rigidity (resistance to change), electrical, optical and
magnetic properties of various media, etc. In classical physics
radiated heat and light do not involve particles or trajectories of
particles.

Philosophers of science who think classical physics was only about
trajectories of particles are forgetting a huge amount of classical
physics.

A field of force is not just a geometrical structure involving
shape: it is a vast collection of different causal powers, (often
continuously varying in space and/or time); including causes waiting
to produce effects if certain conditions occur (if a particle were
here it would be accelerated in that direction, light waves passing
here would be slowed down, or speeded up, or deflected, etc., a
voltage V between these two points would produce a current of I amps
here.)

Moreover as causes produce effects various feedback loops can cause
variations in the original distribution of causal powers (e.g.
oscillating fields, transformations back and forth between kinetic
and potential energy in a pendulum, or between kinetic and elastic
energy in a watch flywheel, etc.)

As waves propagate through fields of force there are very abstract
changes in causal powers located at various portions of space.
Interference between superimposed changing fields can cause changing
distributions of causal powers on different time-scales, without any
motion of particles (at least in classical physics.)

This fact, that the causal powers of complex configurations of
physical entities keep changing as a result of the operation of a
subset of the causal powers, is the basis of vast numbers of
machines, toys, and of course, computers, as well as many natural
phenomena (astronomical processes, biological processes, brain
processes, etc.).

It's the basis of Norbert Wiener's cybernetic theories.

It's crucial to biological organisms, and the processes of
evolution.

And none of this depends on differences between classical and
quantum physics (at this level of abstraction: obviously quantum
phenomena play some role in chemical reactions all round the body,
and a retina can detect individual quanta, which would not exist in
a CP world.)

[HPS]
> namely the idealized zero-size limit of
> small regions of the kind that we can actually observe, namely points,
> and the similar idealized limits of visible lines. Since these basic
> geometric elements are idealizations of shapes and regions of the kind we
> can experiences, one had a logical basis for linking the idealized
> theoretical constructs to real and imagined experiences.

That's just one (empiricist) philosophical view of classical
physics. It's not the only one, and its probably incorrect, unless
the notion of idealisation is construed differently from the way I
understand it.

E.g. I can imagine (i.e. think about, hypothesise) a line going off
to infinity, but I cannot imagine EXPERIENCING a line going off to
infinity. I have no idea what such an experience would be like. Or
rather I can, provided it is something like a line seen going away
horizontally in front of me to the horizon. But then what I
experience is finite. I interpret it as an infinite line, but in
doing so I am using intellectual powers that go beyond the
experience.

Likewise I can imagine (think about) a Euclidean point, but I cannot
experience one.

I think you are committed to a particular philosophy of science
which was once accepted by empiricist philosophers, but many
philosophers of science now reject (including me). On that theory
all concepts used in scientific theories are somehow abstracted from
experiences of instances of the concepts. (Concept empiricism.)

Most deep developments in science are totally unlike that: they
involve postulated ontologies which go beyond anything that has been
or could be experienced. Then the theory introduces links with
previous theories, and eventually there are some links with things
we can observe and measure. But the links may come later, and are
not definitional and they can be very indirect. (I discussed this at
some length in chapter 2 of The (out of print) Computer Revolution
in Philosophy (1978). The position is in some ways close to Karl
Popper's philosophy of science, but differs in many details.)

All this suggests that constantly mixing facts about geometry and
micro-physics with references to "our experience" is just based on a
mistaken philosophy of science, and this leads to conflation of the
question how micro-physics can explain macro-physics with the
question how micro-physics can explain the experience of macro
phenomena.

[HPS]
> The geometric aspect of our experience is  woven into a much richer fabric:
> it does not stand isolated and apart from the rest.

And that's true even if we restrict ourselves to the contents of
classical physics: it has a rich fabric of causal properties, causal
interactions, etc.

> Yet we can say that
> the tide is higher today at noon than it was yesterday at noon; or that the
> moon is higher in the sky tonight at midnight than it was last night at
> midnight etc.: conditions on relative locations can be specified.
>
> Of course, these descriptions are not infinitely precise.

The descriptions can be infinitely precise ("it's exactly 2.35 cm
long").

But I suspect the reality cannot be.

[HPS]
> ....
> The specifications must, of course, fit the characteristics features of the
> phenomena being described. The geometric properties used to specify
> the pertinent physical properties of hurricanes and airplanes and
> locomotives will be different:
> ...
> .....But the only features of
> phenomena that are tied in this way to the micro-physical facts,
> as represented in CPT, are geometric features. Everything else gets
                             ^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^
> left out: the nongeometric features of phenomena are woven onto our knowings
  ^^^^^^^^^
> in intricate and important ways, but, they lack the connection to
> the basic CPT geometric microphysical elements  that arises from the fact
> that those micro elements are idealizations of the macro-geometric
> aspects of our knowings

Apart from the the fact that the geomtry isn't essentially concerned
with knowings these paragraphs also suffer from excessive emphasis
on geometry to the exclusion of other important parts of classical
physics.

[HPS]
> The sense in which certain microphysical facts can entail certain
> macro- geometric facts is this: Suppose Y is a statement that specifies
> certain conditions on some macro-geometric properties that characterize
> observable features of some macro system. Suppose X be a statement that
> asserts that the physical world, as described in CPT, satisfies a certain
> set of microconditions. Then X entails Y if and only if no possible CPT
                               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> world can satisfy X and not satisfy Y.
  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

I guess we are all agreed that micro-physical facts can logically
(mathematically) entail what I called "pure" macro-physical facts.
But the formulation underlined above is incorrect, for it does not
specify what sort of range of possibilities is in question. If you
insert "logically" before "possible" you get an acceptable
formulation.

But that then leaves open the possibility of other analogous, but
non-logical links/entailments. And causal entailment would be such a
case. Then "no possible CPT world" would mean "no world allowed by
the constraints of classical physics".

Or more simply: X entails Y in this sense if whatever makes X true
in a situation will cause things to be the case that make Y true.

> CPT includes not only the basic micro-physical descriptions: it includes, as
> any physical theory must, also the rules for how these descriptions are linked
> to our observations. Otherwise the theory would be devoid of physical content.

No. The physical theory need not in itself say anything about
connections with human experiences or observations. In general that
requires additional/auxilliary hypotheses. That's not part of the
theory.

When I talk about genes I can do so without having any well defined
ideas about how I would observe states and properties of genes. When
I say that a computer is searching for a proof of Fermat's last
theorem I may not know of any way of checking that fact. I may be
too ignorant to decompile its machine code.

> The rules of interpretation of CPT are based in the fact that
                                     ^^^^^^^^^^^^
> the micro-concepts of the theory are idealizations of macro-geometric
> aspects of our human experiences,

No. At most a subset are.

> and the principle that large aggregations
> of microscopic elements can define observable macroscopic shapes.
>
> The painfulness of a pain-experience is not characterized by a shape.

What it is characterised by is the causal power of a painful state
to hold attention and to cause and maintain a desire to end that
state.

By assembling mechanisms in which an appropriate collection of
causal powers can be brought into existence, we can thereby bring
about pain.

> Nor is
> the redness of a conscious red-experience.

Having such an experience is also a state rich in causal powers. Its
phenomenology is far more subtle than that of pain.

I've talked about this before, so won't repeat it in detail except
to say that the experience of a patch of red surface is like the
existence of a field in classical physics: there's a space-filled
collection of dispositions of various kinds. E.g. a red surface can
split in two, allowing another colour to come between the two. It
allows motion across the surface, etc. It allows various possible
changes of shape of the boundary of the patch. And lots more.

So we need to look for mechanisms which can support the maintenance
of such fields of causal powers.

> Human experience has dimensions
> or aspects that are not shape-like.

yes. So has classical physics.

> ...
> Since the nongeometric aspects of our knowings do not have this conceptual
> link to the microphysical facts there will be no comparable entailment
                                                ^^^^^^^^^^^^^^^^^^^^^^
> of nongeometric features from the micro-physical facts, insofar as the
> analysis is confine within the framework of CPT.

But the non-geometric (causal) aspects of our experiences, knowings,
etc. can have links to the non-geometric (causal) aspects of the
micro physics.

These links will not necessarily be "logical entailment" links.
They may be causal links.

> So there is, within the framework of classical physical theory,
> a ``principled distinction'' between the geometric and nongeometric features
> of human experience. This principled distinction allows statements that
> place conditions on microscopic facts to entail such things as
> hurricane-facts and locomotive-facts, yet be unable to entail
> redness-facts or painfulness-facts.

Well, by falsely claiming that classical physics involves nothing
but geometry, you have made it too easy for me to refute the
argument that it cannot account for experiences.

An argument based on a false premiss is unsound.

But of course, I still have the hard work to do: showing what the
causal constitutions of experiences are, and how these relate to the
causal powers of physical systems, which may or may not be
classical. It's ongoing research.


More later on causation and experience. maybe.

Must sleep for now. I hope my sleepiness has not caused too much
sloppiness in my thinking, spelling, sentence construction, etc.

Aaron


From Aaron Sloman Thu Nov 18 10:31:02 GMT 1999
To: Henry
Subject: Re: Reply to Henry's reply  [Part 4] (Back to logic vs causality (and Dave))

This is the last part of my reply to Henry's message
    Date: Sat, 6 Nov 1999
    Subject: Reply to Sloman and Hayes

In the last part of his message Henry gave a summary of some things
I had previously written. I'll make some comments on the unclarity
of some aspects of his summary which leads to mis-placed criticism
of my search for causal rather than logical connections.

I point out some interesting similarities and differences between
what I say and what Dave Chalmers says.

In clarifying Henry's claims I try to enlarge a little on what it
might mean to use logic to formulate causal relations. It requires
the use of a modal operator. And how to give such operators adequate
semantics remains very unclear.

I'll give a tentative example of a case where mathematically
equivalent implementations are causally different i.e. in their
reliability, a concept that refers to "what might happen if" rather
than actual behaviour.

I'll end with a response to Henry's very strange claim that I've
abandoned goals of Western science, and show how if I wanted to do
science his way I could introduce mentalistic terminology into
classical physics, to achieve HIS type of goal.

Henry's summary

> Aaron (Sept 5) asserts that propositions about mental states are
> LOGICALLY disconnected from propositions of physics. This is in line
> with the conclusion that I obtained above. But I obtained my conclusion
> ONLY within the framework of the provably false CPT!
>
> Aaron (Sept 5) suggests that, inspite of this LOGICAL disconnect, the
> two realms are CAUSALLY connected.
> [Hence his title: logical vs causal connections.]
>
> This raises the key issue: Should an adequate physical theory express the
                                                                ^^^^^^^^^^^
> CAUSAL structure of nature in logico-mathematical terms?
  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

I am not sure what the underlined bit means. If it means should we
try to use logic and mathematics wherever possible to express
generalisations, formulate descriptions of particular situations,
derive consequences from our theories, then the answer is surely
yes: I am not attacking the role of mathematics in science

Note, however, that if not all mathematics is logic, then not all of
the mathematical formulations and derivations will be logical) The
reduction of arithmetic to logic was an aim of Frege, though he
thought, like Kant, that geometry could not be so reduced. Whether
all mathematics is reducible to logic remains an open question,
partly depending on how you define "logic". But that's a digression.
I doubt that Henry wants to exclude non-logical parts of mathematics
from science, if there are any non-logical parts.

[HPS]
> The aim of CPT was certainly to do this, at least part. However, it was
> deficient from the outset, because it did not include our conscious
> experiences within the causal structure.

And neither did it include social, legal, economic, biological or
other processes. Some people over-impressed by the achievements of
Newtonian physics certainly thought that in some sense everything
could be reduced to Newtonian physics, but what exactly they *meant*
by that was never very clear.

CPT is not deficient if the phenomena it describes are described
adequately. In Aristotle's physics there were mentalistic terms.
Getting rid of them was an advance, not a deficiency. (Why?)

Many think the same is true of Darwin's theory of evolution: it got
rid of purposiveness and regard that as an advance.

I am not so sure that was correct because once intelligent agents
start selecting mates there could be biological processes that
implement teleological mechanisms in selection, as also happens when
farmers breed animals. But that's a digression.

> And it ultimately failed to fit
> the physical facts.

Agreed.

But there are plenty of machines around whose behaviour is as well
understood from the standpoint of the (false) classical physics as
from the standpoint of quantum mechanics.

What is in question is whether some of the physical machines capable
of supporting mental processes are in that category.

> Aaron's approach is just the opposite of Crick's and the identity theorists,
> who want to say that consciousness is `nothing but' physical brain activity:

yes.

> Aaron's idea is rather to accept the idea, exhibited by CPT, that there
> is a LOGICAL disconnect between mental and physical aspects of nature,
> but to assert that there is a CAUSAL connection between these two aspects
> that is not expressible as a LOGICAL connection..

But a non-physical architecture implemented in physical machinery
will have properties from which we can logically/mathematically
derive other properties. And searching for an architecture with
suitable explanatory powers has to take account of that.

The non-logical causal relations between the ontological levels are
not mysterious. Such things are commonplace and intuitively well
understood in simple cases. We just don't know what the
architectures are in the more complex cases.

Interestingly what I have said about ALL aspects of mentality is
very close to what Dave says about the particular aspect that he
calls "experience", which he contrasts with the functional mental
states, processes, events which he calls "awareness".

It's not clear to me whether he thinks the latter are logically
connected with the underlying physical states, processes and events.
Certainly he suggests that *that* connection is unproblematic,
although details remain to be worked out and it may take a very long
time.

Returning to Henry:
> But why abandon the effort to formulate this causal connection in logical
> terms.

I don't abandon the use of logical formulations. But formulation and
derivation are different.

I do abandon the attempt to do what is provably logically
impossible, namely:

    (A) Given propositions P1, P2, P3,... formulated using only
        concepts of physics, use logic to derive proposition M1,
        which essentially uses mental concepts that are not
        *definable* in terms of those of physics.

(A) as it stands needs to be spelled out more precisely to express
what I am saying is logically impossible. Otherwise there are
trivial counter-examples.

E.g. it is obvious that if Pa is a proposition of physics, and Ma is
a proposition using mental concepts, and if P1, P2, ...Pn logically
entail not(Pa), then they also logically entail not(Pa & Ma).

So in that sense SOME propositions using concepts not definable in
terms of physical concepts ARE derivable from propositions using
only concepts of physics. But in that derivation, Ma and the
concepts it involves are not used essentially, for Ma could be
replaced by any other proposition whatsoever, and the derivation
would still be valid.

I have not attempted to work out a precise and general specification
of what I mean by "essentially uses", except to note that whether
the use is essential or not can vary according to context.

E.g. In another derivation Ma and its components might be used
essentially in the proof of not(Pa & Ma), for example in an argument
of the form:

        X -> not(Pa)
        Y -> not(Ma)
        X or Y
    therefore not(Pa and Ma)

I apologise for not making these details clear in any of my previous
messages. I had not thought about how to make the point in its most
general form with perfect precision, and I probably still have not
done so. But I don't think these qualifications affect any of the
differences between Henry and me.

So the answer to Henry's question
    "But why abandon the effort to formulate this causal connection
    in logical terms"

is that if he means "derive it using logic", then the impossibility
of (A) means that if you don't abandon the effort then you are
trying to use logic to do what is logically impossible.

Of course, if he means merely that we should not abandon
*formulating* our scientific discoveries using logic and
mathematics, then I absolutely agree.

But I think he meant something stronger than "formulate", namely
"prove" or "derive".

[HPS]
> Maybe a more complex logical structure will do..

Logical structure for what? If I discover that whenever I decide to
move my arm something changes in my brain, then this very weak
discovery of a causal connection can certainly be *formulated* using
predicate calculus with a few relation systems and the universal
quantifier.

    For all events m, if m is a decision to move my arm, and ....
    then there is an event p, such that p is a physical change in
    my brain, such that m causally entails p.

It's easy to translate that into some sort of logical notation,
provided that we have a modal operator to express "causally entails"

Some people may think that predicate calculus suffices to express
*causal* conditions, without the addition of a modal operator of
some kind, but I think not. I won't address the issue here. The
important point is what the semantics of such a modal operator would
be.

What sort of modal extension to predicate logic is needed to express
causal relations is an open question (as far as I know).

I doubt that any modal operator specified so far suffices to express
all the types of causal relations we want to express in our
pre-theoretical, or scientific, talk of causes and effects.

For instance, many modal logics use a semantics which depends on the
assumption that there are discrete and complete possible worlds
(complete possible histories of the universe) with a neighbourhood
relation such that if use use "causes" or "necessitates" as our
modal operator then

    "E1 causes (or necessitates) event E2"

means something like

    "For every world w in which E1 is true and which is also a
    neighbour of *this* world (i.e. the actual history of the
    universe) E2 is also be true".

I have no doubt that one can define some interesting modal operators
in that sort of way and use them to express some interesting kinds
of statements about one thing necessitating another.

However, I don't believe that that particular kind of semantics for
modal logic captures our intuitive ideas about causation, including
the ideas that pervade classical physics, or our thinking about
causation of or by mental events.

For instance, when I say that a particular event causes another
particular event, e.g. a collision causes a ball to start moving, I
don't believe that I am referring to complete possible histories of
the universe and relationships between them, and formulating some
universally quantified generalisation over such "possible worlds".

All sorts of people have made statements about causal connections
who have never heard of or thought about possible worlds. A more
plausible semantics for a modal operator expressing causal
relationships might involve not complete possible worlds but
possible *fragments* of the universe, in which various things do and
do not happen.

However I believe it is very difficult to make this notion precise,
partly because there are so many different ways of fragmenting the
universe, corresponding to different ontological levels of
description. This complexity is what I referred to previously as
"the topology of causation".

I think it just is not yet understood: the hardest unsolved problem in
philosophy.
[My paper on "Actual Possibilities" in the proceedings of KR96, the
conference on Knowledge Representation 1996, says more about all
this. It's in the Cogaff web site:
    http://www.cs.bham.ac.uk/research/cogaff/
]

So, in short it's not at all clear WHAT I am saying when I say that
my having a thought causes some physiological change. Expressing it
in some logical notation using quantifiers and a modal operator may
look impressive and may provide a basis for drawing conclusions
precisely, but until the semantics of such statements has been
clarified, it is not going to be at all obvious how one can
logically *derive* such statements from other statements, nor what
one can from them using only logic.

In other words, formulating causal connections using logical or
mathematical notations is easy, and often useful. But it doesn't
in itself help with any of our problems.

However, the successful and (generally) reliable implementation of
many interesting virtual machines in physical machines where facts
about the virtual machines are not logically derivable from facts
about the physical machines gives an existence proof of the
possibility that there are physical phenomena which are causally
related to non-reducible non-physical phenomena.

SELF-MONITORING MACHINES

Another point of interest here is whether the existence of a virtual
machine is simply a matter of how we interpret a process, or whether
the machine itself detects and uses virtual machine events.

Consider an old-fashioned mechanical calculator whose operation
implements addition and multiplication of numbers. We then have a
type of virtual machine containing what could be called
"arithmetical events". Although we treat the events as involving
addition, or multiplication of numbers, its not clear that anything
in the machine itself does so.

By contrast, in a programmable computer, numerical calculations may
be used BY the computer, e.g. counting the numbers of items in two
lists L1, and L2, in order to determine which list is shorter, which
can then be the basis of a decision as to whether L1 should or
should not precede L2 in a bigger list containing both of them.

In a very simple calculator, the events in the arithmetical virtual
machine are recognized and used only by humans e.g. when we look to
see which digits on rotating wheels are visible through holes in the
cover, at the end of a calculation.

By contrast, an important fact about computers is that events in the
virtual machine are used by the virtual machine in conditional
instructions: IF ... THEN ... ELSEIF .... THEN .... etc. The virtual
machine is taking decisions on the basis of inspection of states in
the virtual machine. Likewise its actions change the virtual
machine.

Note that even in standard computers, which even Henry things can be
built on classical principles, we find phenomena appropriately
described as "detecting", "testing", "selecting". This is a small
step towards what we are looking for.

In more sophisticated cases the machine has quite deep information
about the actions available to it: e.g. it can assemble new
sequences of possible virtual machine actions, examine their
properties, select one of the sequences and then execute it.

In more complex cases the machine can observe its own performance,
evaluate the performance, and find ways of improving various aspects
of its behaviour by re-programming itself.

In some cases it can detect attempts by external intruders to
interfere with its normal functioning and try to resist those
attempts.

And so on -- we have only scratched the surface of the space of
possible architectures for virtual machines with increasingly
sophisticated mental powers, and we don't have very clear ideas of
what is and is not possible, especially when multiple computers are
connected.

Can Henry, or anyone, start from a complete physical description of
the micro-structure of a computer and the physical events that occur
at that level and *logically* deduce a statement that the machine
used detection of one of its own virtual machine states to select
between two virtual machine actions?

Not if the language describing the contents of what is detected and
words like words "detect" and "select" are not definable in terms of
physical concepts.

(Of course the virtual machine events they refer to are
*implemented* in physical mechanisms. Implementations are not
definitions, since the same thing can be implemented in many
different ways.)

MATHEMATICAL vs CAUSAL EQUIVALENCE

It is significant that sometimes two physical configurations which
from a certain mathematical viewpoint are identical, can be
different in their causal powers. Thus what follows mathematically
from their description cannot include those causal powers.

I think what follows is an example, though I may not have thought it
through properly (I thank Brian Logan whose criticisms of an
initial formulation helpd be improve this):

Consider engineers who try to make a space ship more reliable by
replacing every computer with three computers, and implementing a
voting mechanism so that what is done at every stage depends on the
majority vote. I assume we all find it obvious that under reasonable
assumptions (what assumptions) that is going be a more reliable
system than one in which the three computers are merely *simulated*
in one single computer with three times as much memory running three
times as fast.

If the simulation is accurate to the level at which the computers
are specified by the engineers, then the behaviour of the three
machines is mathematically equivalent to the behaviour of one
machine simulating all of them.

Every function that can be computed by N turing machines running in
parallel, in synchrony, can be computed by 1 turing machine
simulating the N machines. The situation is different where the
machines are not synchronised, but ignore that for now: we are
assuming synchronised machines which vote whenever they complete a
step.

So how come we know, and engineers use, the fact, that the three
separate machines will be more reliable?

It has to do with the fact that a physical implementation always
includes some unknown elements (either intrinsically or
extrinsically in its environment, possibly the very remote
environment) which might interfere with its intended functioning.

Examples include metal-fatigue, slightly different rates of wear
caused perhaps by slightly different ambient temperatures,
bombardment with radiation or sub-atomic particles coming from
very far away, chaotic consequences of infinitesimal differences in
non-linear systems, or other things we cannot yet even conceive of
etc.

I.e. we don't know, and can never know, all the boundary conditions
involved in any particular implementation. So we could not formulate
them logically.

Yet it seems clear that for very well engineered machines, the
probability that two or more will go wrong in a given time interval
is much lower than the probability that only one will go wrong (if
that is very low). Hence the triple computer solution will generally
be more reliable, even though everything we can specify
mathematically about the functionality of the three-computer system
is identical with what we can specify about the triple simulation
running on one computer.

I.e. if you think there is a mathematical difference, then use it
to change the specification of the triple simulation so that that
difference disappears. Then I claim the three computer system
remains more reliable nevertheless. You cannot redefine the
simulation so as to make the total implementation using that
simulation more reliable!

I don't think any of this depends on whether the machines are made
using quantum or classical physical principles.

Why is the three computer design more reliable? Why do they have
different causal powers despite the mathematical equivalence?

When you produce physical implementations you go beyond mathematics:
you embed something in the world, and then it becomes subject to to
possible influences of many kinds that you cannot possibly
completely predict or understand, because (if the machines are
running long enough) the influences could come from arbitrarily
distant parts of the universe.

Worse, your understanding of the particular level of physics so far
investigated may be incomplete, in which case there could be unknown
causes of unreliability very close at hand.

In many situations we are aware that such inaccessibility of all the
"boundary conditions" means that provided that any particular type
of interference or disruption is unlikely to hit more than one
machine at any time, then if one machine is caused to give an
incorrect result there is a high probability that the other two will
still produce the correct result. So the majoring vote will be
correct.

That argument does not apply when the three machins are simulated in
one machine: for if a component shared by all three simulations is
damaged (part of the CPU) then they will all give wrong results.

Is there some logical, or mathematical, proof that three computers
plus a voting system will be more reliable? Whether there is or not,
it's a causal difference between the two cases, but not a mysterious
one. We intuitively understand such notions of reliability even when
we can't prove our intuitions logically or mathematically.

Note that this notion of reliability is applicable even if the two
systems run forever and given exactly the same results. For
reliability is a causal notion referring to what might happen, or
what would happen if. It is inherently counterfactual. The
relatively unreliable system, might, by chance, perform perfectly.

In a way that's an irrelevant digression. I offer it as an example of
the kind of thing we have to be clear about in understanding the nature
of implementations and to what extent the properties of implementations
can ever be fully known or formulated in a way that logically implies
all the interesting properties of the supervenient systems,
including reliability.

[I first brought this up in an attack on Searle's chinese room
argument in a conference in 1985, later published as
    Did Searle attack strong strong or weak strong AI, in
    A.G. Cohn and J.R. Thomas, eds. Artificial Intelligence and Its
    Applications, 1986, Wiley.
It's also in the cogaff directory.]

[HPS]
> Apparently Aaron
> believe that there is some basic inadequacy of logical thinking

That logical thinking cannot do what's logically impossible, is not an
inadequacy.

It's not an inadequacy of logic that you cannot derive the fact that
it is raining from the fact that my car is silvery grey, though you
can use logic to express the conjuction, or even the false
implication.

> that must
> forever prevent the causal connection between the mental and physical aspects
> from being expressed in a logically coherent way?
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

There is a difference between expressing things in logic, and using
logic to prove something. I approve of the use of logical or more
generally mathematical formulations whenever that is possible and
useful, since it often facilitates communication and thinking
(though not always, e.g. when the structure of a problem can be more
easily understood diagrammatically or verbally.)

The phrase "in a logically coherent way" is ambiguous: it has a
clear weak meaning and possibly also a stronger less clear meaning.
In the weak (innocent meaning) it means "expressed without any
logical contradiction".

It seems that Henry intends something stronger, requiring
derivability between parts of what is expressed. That would be a
requirement for something like a formal system whose axioms were not
mutually independent.

I see no reason to aim for this stronger kind of logical coherence
in scientific theories.

Of course, I like unified and non-redundant theories wherever
possible. In fact what I am trying to do is show how to unify mental
processes and physical processes by demonstrating how physical
architectures can provide support for mental architectures in
something like the way they provide support for other kinds of
virtual machine architectures.

I just don't think it makes sense to aim to do this in terms of
logical derivability because statements in one ontology are
generally not logically derivable from statements in another
ontology.

Interestingly what I am trying to do is something Dave assumes can
be done for what he refers to as the functional aspects of mind and
consciousness ("awareness"). But he assumes that doing that leaves
another task undone, namely explaining why experiences occur, and on
that we part company because we disagree on the analysis of what it
is for experiences to occur.

I don't know whether Dave thinks that existence of functional mental
processes in a virtual machine can be *logically* deduced from
descriptions of the underlying physical processes, or whether he merely
thinks there is some other kind of "conceptual entailment" in such
cases, as suggested in some of our earlier correspondence (in which
Gregg seemed to agree with him.)

I presume Dave agrees with me that there is causal necessitation
of the sort I talk about, though I don't know if he has a theory of what
causal necessitation amounts to. (The hardest unsolved problem in
philosophy.)

He doesn't seem to think that that kind of necessitation is as
interesting as the necessitation involved in his "bridging laws"
between physical events and experience.

[HPS]
> This commitment runs
> counter to the aims of Western science, though it is perhaps in line with
> Eastern traditions.

I don't think anything I've said is counter to the aims of science.

Newton thought that changing the distance between two objects caused
the gravitational attraction between them to change in accordance
with the inverse square law. I don't think he thought he could prove
logically that increasing distance would make the force decrease.

People did try (unsuccessfully) to reason about this by analogy with
an inverse square law of reduction in intensity of radiation from a
point hitting a surface, which is mathematically provable given an
assumption about conservation and symmetry, but I don't think
accepting the non-provability of the existence of gravitational
attraction or its precise form involves abandoning the aims of
Western or any other science.

[HPS]
> Quantum theory is, of course, precisely an expression in logico-mathematical
> terms of the causal structure of the interpenetrating physical and causal
> aspects of nature. And it has the virtue of resting securely on thousands of
> nontrivial theoretical and experimental findings.

I am not saying anything against quantum theory itself. I have
strong reason to doubt the appropriateness of the language some
people (e.g. Henry) use in formulating it, i.e. when they use
mentalistic language. (See below, on how to make classical physics
mentalistic.)

[HPS]
> Aaron suggests that I try to avoid the logical disjunction by bringing
> in mentalistic language. But if one is seeking to bring together the
> mental and physical aspects of nature, which CPT split asunder, should one
> not accept the appropriate two languages and show how the propositions in
> these two languages can be interpreted as parts of a coherent
> combined language?

Yes, but if I decided to replace the terms "repulsion" and
"attraction" in classical physics with "liking" and "disliking", I
could claim that CP involves different kinds of liking and
disliking.

I might also say that since the amount of electromagnetic and
gravitational liking and disliking depends on things like velocity
and distance the entities in CP have to know where the things are
that they like and dislike, and in some cases how fast they are
moving. (Not too far from how Aristotle thought about physics???)

On that basis I could start arguing that we can hope to explain
mentality as arising out of machines built on purely classical
principles, because those principles already include elements of
mentality and that would help to achieve the grand coherence of
scientific explanation that is the goal of Western science, or some
such thing.

But I don't think anyone would be convinced by that argument.

I don't think the argument for using mentalistic language in QM is
any better than that.

[HPS]
> One can surely say a great deal about conceptual things, such as the rules
> of chess, without ever trying to establish what the causal connections are
> between what we are thinking about when we think about these things and what
> is going on in our brains. But if we DO want to inquire about these mind-
> brain connections, which is the endeavour that is behind this whole discussion,
> then I see, at present, no  sufficient reason to reject the Western science
> aim to comprehend the causal connection in logico-mathematical terms.
> The failure of the false concepts of CPT is hardly a good reason to abandon
> this worthy goal.

Well since I have not rejected any coherent goals, and I am merely
pointing out that certain things cannot be done, I don't see that
Henry has characterised what I am doing.

Moreover I am trying to achieve the unification in a different way,
by showing that there's a vast realm of non-physical phenomena that
can be shown to be fully implementable in physical phenomena, with
two way causation between them and physical phenomena, provided that
(a) we understand the topology of causation properly, which has yet
to be achieved and
(b) we work out how to design and implement all sorts of
non-physically definable virtual machine architectures embedded in
suitable networks of physical causation.

I don't think that's anti-science.

[HPS]
> In view of the inadequacy for this task of the empirically false  CPT,
> the most reasonable thing to try next is the empirically valid physical
> theory that has replaced it, which in its original and most most natural
> forms brings mental events directly into the causal structure, within
> a coherent logico-mathematical framework.

And I see this as a verbal device which stops people trying to
understand the architectures that we really must understand if we
want to know how minds work, and how mental phenomena are related to
material phenomena.

Whatever else we disagree on you can't say that I have abandoned
major scientific goals which you are working on. Where we differ is
on how to achieve goals we both aim for. I think you are looking in
the wrong place.

Incidentally I don't think and have never said that QM is
irrelevant.

On the countrary I have claimed often that QM already makes kinds of
machines possible that would not be possible in CP, which includes
computers with all the important kinds of physical characteristics
they now have.

And if someone finds out how to build quantum computers as
described by Deutsch and others, they could truly make a huge
difference, by drastically changing the complexity classes
of many hard computational problems that minds have to address.

And maybe, one day, it will turn out that brains are such quantum
computers. That would be truly fascinating, and a great achievement
for QM.

[the end]

Aaron