Reply to Tom B.

CSGnet Archive CSG1993
1

[From Rick Marken (930617.1100)]

Tom Bourbon (930617.0943) --

But, Martin, doesn't your request for more -- I hesitate to use the word --
information about (in the sense of facts about) the output function vitiate
the claim that a modeler can reconstruct the disturbance from the
perceptual signal alone?

I'll take the liberty of answering for Martin: YES.

Will the
parties *please* clarify the rules of the demonstration?

The rules should have been very simple:

1) I send a sequence of 50 p values;

2) Martin, Allan, etc send we back the 50 d values that occurred
at the same time.

Because of my ridiculously liberal (and flamboyant) nature, I "gave
in" to Martin's demands for the reference signal, r, and the
output function (o = k(e)) despite Bill Powers' protests that this
is information that is NOT available to the control system itself.

In April or early may I sent two sets of 50 p values to Allan Randall
and Gary Cziko. I also sent the corresponding d values to Gary
Cziko. Along with the 50 p values I gave the value of r (I think it was
a constant and 0) and the value of k in the output function along with
the fact that it was a pure integrator.

I never got back the 50 d values from anyone (except Gary, who
properly declared the lack of response a victory for the "no
information in perception" crowd).

I think it's clear from the whole go round on reconstructing d that
Martin et al have no intention of giving up the idea that information
in perception is used by control systems to generate the outputs that
control their perceptual inputs. Demos and models and math bounce of
these guys like bullets off of Superman. So I'm willing to concede that
there must be information in the perceptual input to a control system;
it's there because at least two people in the world (Martin and
Allan) can perceive it and say a lot of things about it. It is
apparently useful, however, only as a basis for conversation; it
seems to contribute nothing to our ability to understand or build
control systems. So even though you and I can't detect the information
in the perecptual input to a control system or understand what it is
(I presume you are having the same problems I am) we are apparently OK
as long as all we want to do is understand purposeful behavior and
build models of purposeful systems.

Best

Rick

2

From Tom Bourbon (930617.1253)

[From Rick Marken (930617.1100)]

Tom Bourbon (930617.0943) --

But, Martin, doesn't your request for more -- I hesitate to use the word --
information about (in the sense of facts about) the output function vitiate
the claim that a modeler can reconstruct the disturbance from the
perceptual signal alone?

I'll take the liberty of answering for Martin: YES.

Rick, your answer does not surprise me. By my question, I guess I revealed
my agreement with your interpretation of the significance of Martin's
request. But I am still eager to see Martin's reply.

Will the
parties *please* clarify the rules of the demonstration?

The rules should have been very simple: ...

Apparently my mind-reading of your interpretation of the rules was correct.
And they are precisely the rules needed to resolve the question of whether
one (any one) can reconstruct d from p.

In April or early may I sent two sets of 50 p values to Allan Randall
and Gary Cziko. I also sent the corresponding d values to Gary
Cziko. Along with the 50 p values I gave the value of r (I think it was
a constant and 0) and the value of k in the output function along with
the fact that it was a pure integrator.

Exactly what should be provided -- and then some. You must have been in a
mellow mood that day.

I never got back the 50 d values from anyone (except Gary, who
properly declared the lack of response a victory for the "no
information in perception" crowd).

OK. That was about where my connection to the net became firm. I suppose
the offer still stands, doesn't it? The test was not the kind that has a
time limit. *Any time* someone shows they can reconstruct the 50
values of d from the 50 values of p (even if they must cheat and use the
other facts you provided) would prove the claim. I think that *all* future
discussions about the importance of information theory in PCT should occur
with the fact of the as-yet-unmet conditions of that offer *clearly* stated
up front. (Or even at the other end -- just so everyone remains aware of
the real state of affairs.)

... So even though you and I can't detect the information
in the perecptual input to a control system or understand what it is
(I presume you are having the same problems I am) we are apparently OK
as long as all we want to do is understand purposeful behavior and
build models of purposeful systems.

That's fine by me!

Until later,
  Tom Bourbon

3

[Martin Taylor 930617 18:10]
(Rick Marken 930617.1100)

Rick, please don't start a flame war on this mailing list:

In April or early may I sent two sets of 50 p values to Allan Randall
and Gary Cziko. I also sent the corresponding d values to Gary
Cziko. Along with the 50 p values I gave the value of r (I think it was
a constant and 0) and the value of k in the output function along with
the fact that it was a pure integrator.

The last two facts were not included. We explained why that information
would be needed.

Demos and models and math bounce of
these guys like bullets off of Superman.

Who was it that refused to believe the results of a demonstration, a
mathematical description, and a model that he initially had accepted
as being definitive and that would not give the results Allan and I
predicted it would? Some loose canon out in California, I think.

Shall we leave it at that?

Martin