[From Bill Powers (940917.1640 MDT)]
Avery Andrews (940916.1518 Eastern Oz Time)--
What philosophers say is that 'purpleness' exists, but it 'supervenes
on' light-spectra. This is pretty close to saying that purpleness is a
CEV. For A to supervene on B implies, among other things, that you
can't get a change in A without a change in B, but there may be plenty
of changes in B that produce no change in A (lots of different spectra
are purple).
And from the same B, you can get A0, A1, A2 ... ? This sounds a lot like
solving a problem by naming it. If B is a set of readings from a
spectrograph (say a plot of intensity versus wavelength displayed on a
CRT screen), and A0 is the color person a0 sees, B0 is the color person
b0 sees, and so forth, where does this supervention take place? For each
observer, we are comparing a perceptual report in the form of a neural
signal in the observer with a wavelength-intensity plot perceived by
some observer, maybe the same observer. Neither the plot nor the signal
is in the same physical form as that which we presume to be causing both
the plot and the signal. And we have different signals being derived (in
different people) from the physical situation that leads to the same
plot. The signals could be different even though people agree to call
them by a single name, "purple." They could be the same, but the signals
might differ due to different perceptual weightings, so that different
spectra would lead to the same signal in different people.
Are the philosophers you mention by any chance taking light spectra as
the objective indication of what is really out there?
If we take CEVs/outputs of PIFs one at a time, they are arbitrary, as
has often been pointed out. But creatures organized along the lines
suggested by PCT need what I'll call 'useful suites' of PIFs, such that
the outputs of some of them can be controlled by controlling the
outputs of others. This is a strong constraint, and note that it
resides entirely in the external world.
No, not at all. Look at Rick Marken's spread-sheet model again. The
second level perceptions are arbitrary combinations of the first-level
perceptions, which are arbitrary combinations of the external variables.
Yet perceptions at both levels (actually all three levels) are perfectly
controllable by acting in the usual manner -- by having an effect on
lower-level reference signals, or on external variables.
The only constraint on what is controllable is whether a mostly
continuous and mostly single-valued relationship exists between an
output and its effects on an input. This still leaves an infinity of
different controllable aspects of the same set of environmental
variables, determined completely by the forms of the perceptual
functions. The degrees-of-freedom limit, of course, applies to
perceptions at the same level that are to be controlled independently at
the same time.
There is another kind of constraint, much less direct. The organism
needs to construct perceptions to control such that when they are
controlled at particular reference levels, intrinsic or critical
variables on which reorganization is based are, as a side-effect, kept
near their reference levels. There is no reason to forbid controlling
other variables, or controlling variables in a way that causes intrinsic
error, but if there is intrinsic error the nature of the controlled
perceptions will be reorganized and such variables will have a limited
lifetime as controlled variables.
So what is it that these philosophers are trying to prove with their
concept of supervention? And I repeat, just where is this supervention
supposed to take place?
P.S. How many hours should I add to GMT to get your local time? Is it 9
or 10?
ยทยทยท
---------------------------------------------------------------------
Best,
Bill P.