Communmicative PCT (was Quantum Thoughts)

Re: Communmicative PCT (was Quantum
Thoughts)
[Martin Taylor 2005.05.15]

Back from Italy, going through old mail.

[From Bill Powers (2005.04.18.05643
MDT)]

Rick Marken (2005.04.17.1850) –

do understand that my remarks could
be irritating and that in a more charitable mood I could have put them
more diplomatically.

I guarantee you that that would not have made them less irritating. As
you well know, the irritatingness of what one says is in the listener,
not in the words themselves.

As Bruce G. apparently intended to illustrate, there is more than one
choice of words that will get across any given message. This means
there are more dimensions in a communication than you need to convey
the basic idea. In addition to getting your message across, you can
also indicate your opinion of the intelligence of the listener, the
honesty, the education, the manners, and so forth. All these
communications run in parallel, with many words doing multiple
duty.

Yes, this was the essence of our layered protocol Theory of
communication. I think Bill has pretty neatly summarized quite a few
of the main points of that theory. As I said at the 1993 CSG meeting,
it was the (slow) discovery that LPT was a special case – PCT applied
to communication – that led me to be concerned with the foundations
and potentialities of PCT more generally.

What Bill is talking about in the above paragraph is what Dave
Waugh and I called (not very originally) “multiplexing”. The
inverse, when you use gestures, body language, and words to support
the delivery of one message, is a process to which we gave the name
“diviplexing”. They are the LPT equivalents of the standard
HPCT hierarchic setup in which eqach control system at one level
passes its perceptual signal up to the perceptual input functions of
several at the next higher level, and received its reference signal
from the outputs of several systems at the next higher level.

There’s a lot more about multiplexing and diviplexing and their
connection with syntax and language functions such as anaphora in
“Multiplexing, Diviplexing and the Control of Multimodal
Dialogue” (M.M. Taylor and D.A. Waugh, Chapter 25 in "The
Structure of Multimodal Dialogue II, M.M.taylor, F.Neel, and
D.G.Bouwhuis, Eds., Amsterdam, John Benjamin, 1998) (It doesn’t
reference PCT because the paper was written in 1990 for a workshop in
1991. The book was very late).

…When the communication is public, the
problem for the listener is multiplied, because even if the speaker’s
opinion is not respected or honored, there are probably others whose
opinions matter to or could have repercussions for the listener (for
example, defamation of character), so the listener may experience
errors relative to desires or intentions relating to bystanders who
hear or read what is being said. And again, if others persist in
manipulating the listener, it is likely that the listener will learn
not to let anybody know what the listener desires or values the
most.

There’s something else, too, about the distinction between public
and private. That is the fact (according to LPT) that a dialogue may
have entirely different intended and actual effects on the particpants
and on the bystanders. It is a tenet of LPT that the content of a
message is not discernable to a third party, though a third party
privileged to all the physical components of the two-way communication
may well be able to hypothesise reasonably accurately both the intent
and the effect of a message.

As Bill said…

… All this took place behind a facade
of being immune to effects from others, propped up by a superficial
interpretation of PCT.

That last sentence, by the way, was
arranged to illustrate how we can use words that appear to be aimed at
one person as a way of inducing errors into others who are not the
ostensible target, but who are intended targets nonetheless. What I
was saying was that anyone who thinks that PCT says we can’t affect
others by our words is being superficial. So anyone who doesn’t want
to be thought superficial, and who has been denying responsibility for
what he or she says, will probably experience some error of some kind.
All that is unusual about this sort of off-center sniping is that I
admit my intent.

Or claim to do so. Only by applying the communicative version of
“The Test” could either the intended reader or a lurker
actually determine whether you tell the truth here. At least that’s
what LPT would suggest, and I think general PCT would also say the
same.

That was a nice post, Bill. Sorry it took me a month to see
it.

Martin

[From Bill Powers (2005.05.16.0614 MDT)]

Martin Taylor 2005.05.15

In addition to getting your message across, you can also indicate your opinion of the intelligence of the listener, the honesty, the education, the manners, and so forth. All these communications run in parallel, with many words doing multiple duty.

Yes, this was the essence of our layered protocol Theory of communication. I think Bill has pretty neatly summarized quite a few of the main points of that theory.

I agree. the Layered Protocol Theory was a nice analysis of communication into levels of organization, prior to PCT.

There's something else, too, about the distinction between public and private. That is the fact (according to LPT) that a dialogue may have entirely different intended and actual effects on the particpants and on the bystanders. It is a tenet of LPT that the content of a message is not discernable to a third party, though a third party privileged to all the physical components of the two-way communication may well be able to hypothesise reasonably accurately both the intent and the effect of a message.

"Not discernible" seems to contradict what follows, unless you're saying that the content is discernible with 100% certainty by the two participants. Or else perhaps you don't consider me a participant in this dialog, because if I were I would not be asking for clarification....

Is there a difference between saying a third party is "privileged to all the physical components" and saying the third party receives all the messages sent by both sides at the lowest (physical) level? What you appear to be saying might have some heavy implications for intelligence operations. If your meanings is what I think you mean, intercepting communciations between two other parties is not a very reliable way to discover what they are saying to each other at the higher levels (meanings, implications). Experience of the USA in Iraq would appear to support that idea.

All that is unusual about this sort of off-center sniping is that I admit my intent.

Or claim to do so. Only by applying the communicative version of "The Test" could either the intended reader or a lurker actually determine whether you tell the truth here. At least that's what LPT would suggest, and I think general PCT would also say the same.

*Claim to admit what I said I was admitting? I claim that I just wrote the preceding "sentence" after the *, but unless you want to hypothesize mysterious third parties who add things to my messages, the current sentence is the proof of the claim that I wrote the starred sentence, isn't it? I said something which could have been taken by a third party as a criticism, then wrote that I admitted that I said it and recognized its possible meanings. I was therefore aware of the point to which I was admitting in the same message, and I don't see any possibility that the admission could be spurious. I might spuriously claim, after the fact, that I had no such intention of sniping at bystanders, but how could the claim be spurious when the claim and the utterance are in the same message, and agree?

I'm not completely sure of my logic here. How do you read it?

Best,

Bill P.

Re: Communmicative PCT (was Quantum
Thoughts)

[From Bill Powers (2005.05.16.0614
MDT)]

Martin Taylor 2005.05.15

In addition to getting your message
across, you can also indicate your opinion of the intelligence of the
listener, the honesty, the education, the manners, and so forth. All
these communications run in parallel, with many words doing multiple
duty.

Yes, this was the essence of our layered protocol Theory of
communication. I think Bill has pretty neatly summarized quite a few
of the main points of that theory.

I agree. the Layered Protocol Theory was a nice analysis of
communication into levels of organization, prior to PCT.

There’s something else, too, about the
distinction between public and private. That is the fact (according to
LPT) that a dialogue may have entirely different intended and actual
effects on the particpants and on the bystanders. It is a tenet of LPT
that the content of a message is not discernable to a third party,
though a third party privileged to all the physical components of the
two-way communication may well be able to hypothesise reasonably
accurately both the intent and the effect of a message.

“Not discernible” seems to
contradict what follows, unless you’re saying that the content is
discernible with 100% certainty by the two participants.

I worded it badly, perhaps. The esential point is that the
content of a message is the effect it has on the recipient. In an
overheard collaborative exchange, there are three kinds of
participants: originator, intended recipient, and bystander (note the
word “collaborative”). In such an exchange, the recipient
may know what effect the message is having (i.e. what its current
content is), but may not know what effect the originator intends.
Likewise, the originator knows what effect the message is intended to
have, but not what effect it actually has (so far).

The bystander is in the same position as the recipient, but
differs in that the recipient has the opportunity to perform “The
Test” on the originator, by offering to the originator a version
of the message content (so far), to see whether it agrees with the
originator’s reference version of the message content. If the two
differ, the originator’s relevant control system has an error value
differnt from zero, and further output continues the message until
each party has an error value approaching zero. The recipient is
controlling for perceiving the originator to believe that the message
content was correctly received, while the originator is controlling
for perceiving that it was in fact correctly received.

Whereas the collaborating participants can use control to
converge on the intended message content (the intended effect on the
recipient), the bystander has no opportunity to exercise any such
control, which is what I shorthanded into “not
discernable”.

Or else perhaps you don’t consider
me a participant in this dialog, because if I were I would not be
asking for clarification…

No, if you ask for clarification and I respond, that is what
makes us both be particpants in a collaborative dialogue.

Is there a difference between saying a
third party is “privileged to all the physical components”
and saying the third party receives all the messages sent by both
sides at the lowest (physical) level?

I think those are paraphrases, having the same intent.

What you appear to be saying might
have some heavy implications for intelligence operations. If your
meanings is what I think you mean, intercepting communciations between
two other parties is not a very reliable way to discover what they are
saying to each other at the higher levels (meanings, implications).
Experience of the USA in Iraq would appear to support that
idea.

Yes. An example I used was of a couple arriving early at a party.
As they arrive, the host says to his wife “I’m going for the
beer. Back soon.” The arriving couple sees that as a message of
helpfulness, whereas it is actually the reverse, since he was asked to
set the dinner table just before they arrived. Literally, he has
informed her that he is going for the (necessary) beer, but more
importantly, he has said “The hell with you. You set the bloody
table yourself.”

At a much lower level, in World War II the Navaho code-talkers
allowed the Japanese to hear all the sounds they made, but only the
Navaho at the other end could use those sounds to military
advantage.

All that is unusual about this sort of
off-center sniping is that I admit my intent.

Or claim to do so. Only by applying the communicative version of
“The Test” could either the intended reader or a lurker
actually determine whether you tell the truth here. At least that’s
what LPT would suggest, and I think general PCT would also say the
same.

*Claim to admit what I said I was admitting?

No, what you wrote was a claim that you admit what you wanted the
reader to believe you admitted. Whether what you admitted was your
actual intent is not determinable, not from the one-way statement.
“The Test” might be able to determine what you were
controlling for, both in the admission itself, and in the intent that
was admitted (though I suspect the dynamics of the situation would
make “The Test” hard to apply in practice). Isn’t MOL
designed in part to allow you to discover your own intent?

I claim that I just wrote the preceding
“sentence” after the *, but unless you want to hypothesize
mysterious third parties who add things to my messages, the current
sentence is the proof of the claim that I wrote the starred sentence,
isn’t it?

That you created the letter string, the word sequence, and
anything at those levels, there seems little question. The intent of
writing the starred sentence seems to be clarified by the longish
supplementary explanation, at least at one level. But there are
doubtless other parallel messages, and higher-level intentions also
implicit, and those might be seen by people who know you well, or
elicited in collaborative dialogue by a stranger.

I said something which could have
been taken by a third party as a criticism, then wrote that I admitted
that I said it and recognized its possible meanings. I was therefore
aware of the point to which I was admitting in the same message, and I
don’t see any possibility that the admission could be
spurious.

Now I’ll put in a contentious way something that might be said
more technically:

You bloody well shouldn’t take my technical comment as
disagreement with what you seem to have intended as your message
content when you said “I admit my intent”. I never said you
didn’t, and any fool ought to have seen that I didn’t. I said we (and
you) couldn’t know whether your statement that you admitted your
intent was correct, which is quite different from saying that you
lied. So there! Boo snubs and squishy mess!

I might spuriously claim, after the
fact, that I had no such intention of sniping at bystanders, but how
could the claim be spurious when the claim and the utterance are in
the same message, and agree?

There are lots of ways of improving one’s chances of creating the
desired message content (effect on the recipient). One of them is
redundancy of the kind you mention. I think that in practice, almost
everyone reading what you wrote would be happy to accept that you
correctly stated your intent, and by using the word “admit”,
you indicated something bout your own feelings about your intent. So,
in practice, very little dialogue feedback would be required to ensure
a high probability of correct understanding on the part of the
readers.

I’m not completely sure of my logic here.
How do you read it?

Interestingly.

What I suspect happened was that I deliberately violated a
convention of dialogue, in that I specifically suggested that
clarification might be needed on a point that you had never expected
to need clarification. This violation of convention may have led you
to believe that I failed to receive the intended content, or worse,
that it had conflicted with some previous belief I might have held.
That perception (by you) differed from your reference value for what
effect your message was intended to have on me, so you generated a
suplementary message that (as perceived by me) had two purposes: (1)
to enhance the probability that the original message would be
correctly received, and (2) to delve further into the technical point
I raised, because I had not made it clear.

In LP Theory there is a construct called the “General
Protocol Grammar” or “GPG”. It was derived almost 20
years ago, but survived quite a few evolutionary stages in the
developement of LPT. Its form was originally devised more or less
intuitively, but observations of real dialogues never demanded any
significant changes. It does have a formal representation beynd the
intuitive graph shown below.

However, I’ve never tried to derive it from PCT first
principles, and it might be interesting to make that attempt. The
formal structure uses elements that translate into things like
“The originator believes the recipient believes …” and
“The recipient intends that the originator believe that the
recipient believes …”. At first glance it looks like reference
and perception, and indeed the diagrams also look like that. So the
“subposition” of a PCT infrastructure might be a question of
language only. I don’t know.

For what it’s worth, here’s a node-and-arc version of the GPG.
Nodes labelled “O” mean it’s the originator’s turn, whereas
nodes labelled “R” mean it’s the recipient’s turn. We know
very well that the occupancy of the nodes and arcs is not discrete, as
feedback loops are simultaneously active all the way around. So the
grammar is only a cartoon. But it does seem to describe pretty well
the transmission of one single message from O to R at one protocol
level.

The GPG includes not only these nodes and arcs, but also for each
arc a set of possible implementations (one to four possibilities per
arc).

I hope that this will make sense if you try to follow a message
from the leftmost node (labelled OS) through the various possibilities
to the “End” node. (The arc labelled “E-feedback”
refers to the originator’s initial perceptions of the recipient’s
state; E == “expectation” or something like that). The
“Normal Feedback” arc includes a “Null”
possibility, for the case in which both originator and recipient are
satisfied that the message will be correctly understood, which is
usually the case at the lowest levels, such as the level of the word,
and is often the case at higher levels among lonmg-time friends.

Martin

GPG.JPG

[From Bill Powers (2005.05.16.2008 MDT)]

Martin Taylor (2005.05.16) –

I hope that this will make sense
if you try to follow a message from the leftmost node (labelled OS)
through the various possibilities to the “End” node. (The arc
labelled “E-feedback” refers to the originator’s initial
perceptions of the recipient’s state; E == “expectation” or
something like that). The “Normal Feedback” arc includes a
“Null” possibility, for the case in which both originator and
recipient are satisfied that the message will be correctly understood,
which is usually the case at the lowest levels, such as the level of the
word, and is often the case at higher levels among lonmg-time
friends.

Your diagram needs more explanation than you gave it. Or perhaps I should
admit that I’m not as good at deciphering flow charts as I used to
be.

In my new book, Chapter 5 (which is as far as it’s gone so far) opens
expecting a “live block diagram” of a control system to be
running on the user’s computer, the canonical PCT diagram running at 60
frames per second but appearing to do nothing at first. After the diagram
is shown in the text, this follows:

···

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

If you’re not used to circuit
diagrams or block diagrams of circuits you may not realize what this is.
It is not a diagram of things happening and leading to other things
happening, as many diagrams of this sort are. It is not, in other words,
a flow chart. A flow chart is a way of pictorially representing a
sequence of operations, as in following a recipe to bake a cake. One box
in the flow chart might say “Mix liquid ingredients in a
medium-sized bowl,” with an arrow leading to the next box which
says, “Blend mixture into dry ingredients.” Asking what
the process in the first box is doing after you have gone on to the next
box would be a nonsense question, meaningless. The labels describe
behaviors, not mechanisms that produce those behaviors. They are
instructions, descriptions of what to do, and aren’t doing anything
themselves. It’s your attention that moves from one box to the next. For
those who know the reference, the TOTE unit proposed long ago by Miller,
Galanter, and Pribram and still appearing in the literature is a flow
chart: Test → Operate → Test → Exit describes actions that
a system is doing one at a time, so while “Operate” is going
on, it makes no sense to ask what the blocks labeled “Test” and
“Exit” are doing.

Fig. 5-1, on the other hand, is a system diagram, a block diagram of a
working system. Each box in this diagram represent a specific ongoing
process carried out by a physical structure or mechanism that
continuously converts a changing input into a changing output. The
behavior of the boxes depends on what the variables at their inputs do.
You can therefore ask what one box is doing while another is doing
something else, just as you can ask what the engine of a car is doing
while gas is flowing from the gas tank to the engine and the speedometer
is indicating speed. Such things do not take turns happening; they are
all happening at the same time.

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

It seems to me that
your “GPC” is an attempt to combine a flow chart with a
system diagram, and that the result is confusing. for that reason. I
suppose the sequential nature of speech has something to do with it –
speech happens like a program unfolding and people usually have to take
turns, so only one thing, apparently, is happening at a time, as in a
flow chart.

But maybe that’s not right. I have a goal for perceiving that you
understand me, and that goal is in effect whether I am speaking or you
are speaking. We could be anywhere in the GPC flow chart and that
reference condition would exist in me, as would a perception of the
degree to which I am succeeding, and a degree of error, and an ongoing
process of formulating output which is still going on while you are
speaking (modified, of course, by what I am hearing from you).

Have you considered the problem I’m talking about? The flow chart aspect
of the Layered Protocol idea makes it seem that there is an alternating
series of mutually-exclusive events going on, yet the PCT concept
suggests that control systems in both participants are in operation
simultaneously, all during the interchange. It seems to me that if you
could get the flow-chart aspects separated from the system diagram of the
control systems, everything might get neater. As it is, the flow chart
seems to be mainly about a program that is running – operations
interspersed with choice points – and not so much about the organization
of the system within which the program runs. It certainly makes no sense
to ask, when the process is at a certain node in the GPC, what is
going on in the other nodes. That’s enough to tell us that we’re looking
at a description of sequential behaviors, not a diagram of the structure
of a working system. So what is the structure of the working system
behind the flow-chart of the GPC? Have you ever looked at it that
way?

Best,

Bill P.

Re: Communmicative PCT (was Quantum
Thoughts)
[Martin Taylor 2005.05.16.23.13]

[From Bill Powers (2005.05.16.2008
MDT)]

Martin Taylor (2005.05.16) –

I hope that this will make sense if you
try to follow a message from the leftmost node (labelled OS) through
the various possibilities to the “End” node. (The arc
labelled “E-feedback” refers to the originator’s initial
perceptions of the recipient’s state; E == “expectation” or
something like that). The “Normal Feedback” arc includes a
“Null” possibility, for the case in which both originator
and recipient are satisfied that the message will be correctly
understood, which is usually the case at the lowest levels, such as
the level of the word, and is often the case at higher levels among
lonmg-time friends.

Your diagram needs more explanation than
you gave it…It seems to me
that your “GPC” is an attempt to combine a flow chart
with a system diagram, and that the result is confusing. for that
reason. I suppose the sequential nature of speech has something to do
with it – speech happens like a program unfolding and people usually
have to take turns, so only one thing, apparently, is happening at a
time, as in a flow chart.

But maybe that’s not right. I have a goal for perceiving that you
understand me, and that goal is in effect whether I am speaking or you
are speaking. We could be anywhere in the GPC flow chart and that
reference condition would exist in me, as would a perception of the
degree to which I am succeeding, and a degree of error, and an ongoing
process of formulating output which is still going on while you are
speaking (modified, of course, by what I am hearing from you).

Have you considered the problem I’m talking about?

Yes, we originally (mid 1980’s) were thinking in terms of
sequential dialogue acts, and the activity in the GPG chart moving
from one point to another, but quite early on we recognized the issues
you raise.

The problem is how to express it on paper. It’s not easy, even in
verbose text, and it’s less easy in a static diagram. The best we came
up with is to supplement the diagram with text saying that many arcs
and nodes may be simultaneously active, and that even when there is
only one transition occurring, it occurs progressively, not as an
event in which the activity switches instantaneously from one node to
the next.

If you draw the PCT diagrams, they don’t look at all like the GPG
diagram. You have to show the control systems in each individual,
showing what each is trying to achieve (reference values) and the ways
in which the output of one affects the inputs of another.

The GPG doesn’t do that. Instead, it deals with what happens
when, for example, the recipient’s perception of the originator’s
output is that the originator does (does not) believe that the
recipient correctly understands – in other words, whether or not the
recipient has significant error in his system that controls for the
originator being content with the transmission of the message.
Similarly for many other possible situations in the dialogue. In that
sense, the GPG is a kind of flow chart, but a funny (and fuzzy) one in
that you can seldom point to a place in it and say “the dialogue
is there at this instant”.

The flow chart aspect of the Layered Protocol idea
makes it seem that there is an alternating series of
mutually-exclusive events going on, yet the PCT concept suggests that
control systems in both participants are in operation simultaneously,
all during the interchange. It seems to me that if you could get the
flow-chart aspects separated from the system diagram of the control
systems, everything might get neater.

I did do a system diagram of the apparently necessary control
systems in one paper. It’s quite complex within each participant, even
disregarding the issue that there are different levels of message
abstraction. Within any level, the originator is controlling a
perception of the recipient’s state in respect of the intended message
(a multidimensional state, and hence a multidimensional perception),
as well as a perception of whether the recipient believes the message
to have been received as the originator intended, and (one further
layer of recursion) a perception of whether the recipient perceives
the originator to believe that the message has been received as
intended. Although in principle that recursion could be infinite, we
have never found an instance in which it needs to go beyond the three
layers I mentioned.

As
it is, the flow chart seems to be mainly about a program that is
running – operations interspersed with choice points – and not so
much about the organization of the system within which the program
runs.

Exactly so.

It certainly makes no sense to ask, when the process
is at a certain node in the GPC, what is going on in the other
nodes.

It does, often. Particularly it makes sense to ask what is
happening simultaneously at the two ends of any arc.

The arc represents the effects of one partner’s output on the
other partner. Take for example an arc connecting an “R”
node to an “O” node. Such an arc represents feedback from
the recipient to the originator. The outputs that constitute that
feedback take time, and even though the Recipient stays at the node
until his output is finished for the moment (which it will, provided R
is neither a New Yorker nor a Spaniard), the state of the originator
may fluctuate during the recipient’s output. At some point, perhaps
while the recipient is still talking, the originator may believe that
he has enough to move on to the next stage, and may overlap output
with the recipient’s. If that happens, for a while several nodes and
arcs would be active simultaneously. But it’s hard to diagram that
kind of thing.

That’s enough to tell us that we’re looking at a
description of sequential behaviors, not a diagram of the structure of
a working system. So what is the structure of the working system
behind the flow-chart of the GPC? Have you ever looked at it that
way?

I can’t remember for what
paper or presentation I drew the system diagram – it could have been
for the LPT papers in the PCT special issue of IJHCS. If I find it,
I’ll try to send it if I can find an example anywhere. It was rather
more complicated than I like to remember! Another thing we tried, but
I don’t think we included it in a published paper, was to do dynamic
graphs of the “occupancy” of nodes during the dialogue. I
can’t remember that effort being very successful, but we wanted to do
it as a possible step toward automating a user-friendly
interface.

Martin

[From Bill Powers (2005.05.18.0444 MDT)]

Martin Taylor 2005.05.16.23.13 –

Have you considered the
problem I’m talking about?

Yes, we originally (mid 1980’s) were thinking in terms of sequential
dialogue acts, and the activity in the GPG chart moving from one point to
another, but quite early on we recognized the issues you raise.

The problem is how to express it on paper. It’s not easy, even in verbose
text, and it’s less easy in a static diagram. The best we came up with is
to supplement the diagram with text saying that many arcs and nodes may
be simultaneously active, and that even when there is only one transition
occurring, it occurs progressively, not as an event in which the activity
switches instantaneously from one node to the next.

This is the problem with trying to diagram the behavior rather than the
system that is behaving. The actual behavior will depend dynamically on
how the interaction goes – even I can’t diagram the behavior of the
participants in my Crowd program, because the details vary (there is some
built-in random dither), and I suspect that the relationships can become
chaotic, in the sense of being unpredictable even in principle. The
butterfly effect. The more nearly head-on an incipient collision is, the
less predictable is the choice of avoiding it by turning to the left or
the right, for example.

What keeps a dialog organized is the intention of each party involved and
the effort to keep errors small. It would be a real tour de force if you
could come up with a communication control system – even at just one
level – that would actually generate communications toward the end of
achieving and maintaining certain goal-states in the conversation, or
nonverbal states. A generative model, that is, as opposed to a
descriptive model – a system design such that two of them would actually
generate a dialog which you could then chart as a flow diagram. It might
be necessary to develop a limited vocabulary of words designating
nonverbal things that the parties are trying to control by using the
words on each other.

I know I’m describing a huge and long-term effort. I don’t really expect
you to abandon what you’re doing to carry it out.

Best,

Bill P.

Re: Communmicative PCT (was Quantum
Thoughts)

[From Bill Powers (2005.05.18.0444
MDT)]

Martin Taylor 2005.05.16.23.13 –

Have you considered the problem I’m talking
about?

Yes, we originally (mid 1980’s) were thinking in terms of sequential
dialogue acts, and the activity in the GPG chart moving from one point
to another, but quite early on we recognized the issues you raise.

The problem is how to express it on paper. It’s not easy, even in
verbose text, and it’s less easy in a static diagram. The best we came
up with is to supplement the diagram with text saying that many arcs
and nodes may be simultaneously active, and that even when there is
only one transition occurring, it occurs progressively, not as an
event in which the activity switches instantaneously from one node to
the next.

This is the problem with trying to
diagram the behavior rather than the system that is behaving. The
actual behavior will depend dynamically on how the interaction
goes…

What keeps a dialog organized is the intention of each party involved
and the effort to keep errors small.

Yes, that’s the underlying mechanism. But I think you are too
quick to dismiss the possibility of describing the likely behavioural
paths (note: “likely” not “possible”) in a
collaborative interaction. If the interaction isn’t collaborative, but
includes conflicting goals, then the problem becomes much less
well-defined. In a collaborative dialogue, the GPG seems to provide a
pretty reasonable cartoon description of the overt behaviour that
actually happens.

In the development of LPT (before its PCT-substrate was
recognized), the GPG developed from two sides: an intuition about what
one might do under different conditions of the state of communicating
the current message (at one level), and informal analysis of real
dialogues to see whether they took any turns that were not already
incorporated in the GPG. In retrospect, the intuitive analysis always
depended on the formula: “If I were in this situation, and wanted
that situation, what would I do.” In other words, asking about
the likely kinds of error that might be leading to dialogue output
to control a current perception. The observational approach was, of
course, subject to the qualification that, as third parties, we could
never be assured of our interpretations. The two together seem to
provide a consisten formulation, though.

In formalizing the grammar, the diagrammed GPG was ignored.
Instead, we started with an invented formal notation based on three
statements of fact (“real world” conditions unknowable to
the participants). The three statements were:

P1: The recipient has made an interpretation of the primal
message

P2: Whenever P1 is true, then the recipient’s interpretation will
be adequate

P3: It is not worth continuing to try to reach the state (P1 &
P2).

(quoted from Taylor and Waugh, “Layered Protocol Analysis of
Dialogue” in “Abduction, Belief and Context in Dialogue:
Studies in Computational Pragmatics, H. Bunt and W. Black, Eds,
Amsterdam, John Benjamin, 2000”). (Note: the “primal
message” is the originator’s reference for the state the
recipient should be in after the message has been correctly
interpreted)

Each participant (and any onlooker) may perceive any or all
of these propositions to be true, false, or indeterminate to some
degree. In other words, belief in their truth is a continuous (fuzzy,
if you like) variable.

If, for example, in a collaborative dialogue the originator
strongly believes P1 to be false, it will not matter to the originator
what interpretation the recipient believes she has made, but what the
originator does about it will depend on whether the originator
believes the recipient to believe P1 to be true – as well as whether
the recipient believes P2 if the originator believes the recipient
believes P1.

In a collaborative dialogue, both parties have reference levels
that all three propositions be simultaneously true. In a
non-collaborative dialogue, the reference levels for the two parties
may differ in respect of any or all of the three propositions. The
reason it is plausible to use a process model to regenerate the GPG
for a collaborative dialogue is that the reference values for all
propositions are assumed to be known, as are the possible mechanisms
available for output.

To explain this last, a mechanism for the example condition above
is to send a (lower-level) message that informs the recipient that the
originator does not believe the recipient to have made an
interpretation. The means by which that lower-level supporting message
is implemented is a question for the GPG at that lower level, just as
in the normal PCT hierarchy, the mechanism is unspecified whereby an
output of a higher-level control system affects its perceptual
variable.

Continuing with the formalization…

We recognize three variables connected with each proposition, all
of which take on a continuum of values that you might say range from
-1 to +1, or from minus to plus infinity. Let’s use -1 to +1 for
convenience. The variable correspond to the three main variables of an
elementary control unit (ECU): the perceptual signal, the reference
signal and the output gain.

Variable 1 is the degree of belief in the proposition (a
perception). Variable 2 is a goal for the belief in the perception (a
reference, which in a collaborative dialogue is +1). Variable 3 is the
intention to produce a change in the perception (the output gain). We
simplified the analyses by assuming the output gain to be constant, so
neither it nor the reference signals actually appear very much in the
formalism.

The notation we used was not very transparent, but it was the
best we could come up with. If participant A has a degree of belief x
about proposition P, we wrote x = A(b,P). For a goal, we substitute
“g” for “b”, and for an intention, we used
“i” for “b”. Since neither “i” nor
“g” are much used except for exposition, we usually elided
the “b” and wrote x = A(P).

The example above would be written “0 > Originator(P1)”
– zero is greater than the degree of belief the originator has that
the recipient has made an interpretation. If the Originator believes
also that the recipient believes that the recipient has made an
interpretation, we get

0 > Originator(P1) & 0 <
Originator(Recipient(P1))

The second term represents another perception that must be
controlled for the dialogue to progress in a collaborative way.
Indeed, as I mentioned in my previous message, we found that one (and
only one) more recursion is needed. the Originator needs to control
(b,O(b,R(b,O(P))) using O and R for Originator and Recipient of the
primal message. The revipient has a corresponding set of belief values
to control.

Given that there are three fundamental propositions and three
recursion levels (including the zeroth level of belief about the
proposition), each participant is controlling nine perceptions. The
values of these nine perceptions defines a point in a 9-D space, and
different regions of the space map onto the nodes of the GPG – at
least those that are likely to be occupied in a collaborative dialogue
do. They determine also the implementations of the arcs that connect
the nodes. If I remember correctly, when we itemized the
implementations of th arcs (including “do nothing because
perception matches reference”) we counted someting like 47
different plausible things that might happen at one dialogue level in
communicating one primal message.

Some, of course, are more likely than others. At the word level,
between people who talk the same dialect of the same language, and are
talking in a quiet environment, the primal message (the word) is
almost always going to be correctly interpreted (fact), and the
participants are likely to beieve that to be so. So, at O1, where the
word is delivered, the originator immediately believes P1, P2, and P3,
and that the recipient believes the three propositions, and that the
recipient believes the originator believes them. Knowing (or
believing) all that, the originator can, legitimately and
collaboratively, proceed immediately to the next word. Seldom will the
Recipient interrupt with a feedback message asking for clarification
at that (the word) level. So at that level, almost all of the 47
possibilities are essentially unused.

However, if the originator intends to explain General Relativity
to a precocious ten-year-old, the sequence of feedback messages may
last several years before the teacher and student both believe P1, P2,
and P3 and the two levels of recursion about each other’s belief
states. All 47 paths in the GPG (other than the “Abort”
paths) are likely to be well used over those year of passing that one
message.

It would be a real tour de force if you
could come up with a communication control system – even at just one
level – that would actually generate communications toward the end of
achieving and maintaining certain goal-states in the conversation, or
nonverbal states.

As you see, we’ve gone at least some way toward specifying such a
model through the formalism. Somewhere or other, I’ve diagrammed at
least a part of it in the visual language of control systems, but I
don’t remember where that was. If I run across the diagram, I’ll post
it. We have never tried to embody the formalism in working code, which
would be a major project, as it would involve not only the control of
nine perceptual variables at each level (no big deal), but also the
implementation of the perceptual input functions (read “language
understanding functions”) and the output functions (read
“language generation systems”).

A generative model, that is, as opposed
to a descriptive model – a system design such that two of them would
actually generate a dialog which you could then chart as a flow
diagram. It might be necessary to develop a limited vocabulary of
words designating nonverbal things that the parties are trying to
control by using the words on each other.

A toy of this kind might be do-able since it would eliminate the
need for the language systems, but it would still be, as you say, a
huge effort. It’s a little like what Allan Randall and I tried to do
years ago by building a “Syntactic Little Baby” (based on
your “Little Baby”) that was supposed to learn to perceive
and control a simple syntax that created a stream of As and Bs. Even
that turned out to be a bigger project than we had money for.

I know I’m describing a huge and
long-term effort. I don’t really expect you to abandon what you’re
doing to carry it out.

I’m glad of that :slight_smile:

But I hope the above helps people to understand a little of where
LPT fits on top of PCT.

Martin

[From Bill Powers (2005.05.20.1859 MDT)]

Martin Taylor (2005.05.19) –

This is the problem with trying
to diagram the behavior rather than the system that is behaving. The
actual behavior will depend dynamically on how the interaction
goes…

What keeps a dialog organized is the intention of each party involved and
the effort to keep errors small.

Yes, that’s the underlying mechanism. But I think you are too quick to
dismiss the possibility of describing the likely behavioural paths (note:
“likely” not “possible”) in a collaborative
interaction. If the interaction isn’t collaborative, but includes
conflicting goals, then the problem becomes much less well-defined. In a
collaborative dialogue, the GPG seems to provide a pretty reasonable
cartoon description of the overt behaviour that actually
happens.

I didn’t mean to dismiss it, only to note that the description of
behavioral paths (“flow chart”) is an attempt to predict likely
observations on the basis of past experience rather than to derive the
predictions from first principles (“first” inside PCT anyway).
If you have a correct system diagram, then no observation should ever
contradict it. With flow charts, all you can say is that some or most do
it this way, and some don’t. It’s like trying to predict tracking
behavior in the sense of trying to predict the reference settings,
disturbances, and parameters in a given situation. You can do that only
by examining a lot of people in relevant experiments.

In the development of LPT
(before its PCT-substrate was recognized), the GPG developed from two
sides: an intuition about what one might do under different conditions of
the state of communicating the current message (at one level), and
informal analysis of real dialogues to see whether they took any turns
that were not already incorporated in the GPG. In retrospect, the
intuitive analysis always depended on the formula: “If I were in
this situation, and wanted that situation, what would I do.” In
other words, asking about the likely kinds of error that might be
leading to dialogue output to control a current perception. The
observational approach was, of course, subject to the qualification that,
as third parties, we could never be assured of our interpretations. The
two together seem to provide a consistent formulation,
though.

Yes. I think that is a good example of an application of PCT, once you
bring the theory into it. But the basic approach without PCT, trying to
fit a flow chart to as many different examples as possible, is empiricism
coupled with statistical generalization, perfectly useful pursuits but
not theoretical.

In formalizing the grammar, the
diagrammed GPG was ignored. Instead, we started with an invented formal
notation based on three statements of fact (“real world”
conditions unknowable to the participants). The three statements
were:

P1: The recipient has made an interpretation of the primal message

P2: Whenever P1 is true, then the recipient’s interpretation will be
adequate

P3: It is not worth continuing to try to reach the state (P1 &
P2).

OK, but that is again an attempt to find a statistical generalization
that covers as much data as possible, not to explain the observations in
terms of an underlying model. This sort of investigation isn’t within my
field of competence, which is to say that I don’t do it much.

…I hope the above helps people
to understand a little of where LPT fits on top of PCT.

I can see that it’s an application of PCT to provide some theoretical
infrastructure for the empirical approach. I’m all for it, as long as I
don’t have to do that too.

Best,

Bill P.

Re: Communmicative PCT (was Quantum
Thoughts)
[Martin Taylor 2005.05.21.13.49]

[From Bill Powers (2005.05.20.1859
MDT)]
Martin Taylor (2005.05.19)

In formalizing the grammar, the
diagrammed GPG was ignored. Instead, we started with an invented
formal notation based on three statements of fact (“real world”
conditions unknowable to the participants). The three statements
were:

P1: The recipient has made an interpretation of the primal message

P2: Whenever P1 is true, then the recipient’s interpretation will be
adequate

P3: It is not worth continuing to try to reach the state (P1 &
P2).

OK, but that is again an attempt to find a statistical generalization
that covers as much data as possible, not to explain the observations
in terms of an underlying model.

I think you misunderstood much of my last message, because you
couldn’t make that statement if you had understood it as I intended.
(Semi-formally, the values of my three direct perceptions at the
moment are that I believe P1, disbelieve P2, and disbelieve P3; the
values of the first recursion perceptions are that I also believe
reasonably strongly that you believe P1, P2, and weakly that you
believe P3).

I have goals (references) that I should believe P1, P2, and P3.
Therefore there is error in my control systems for the values of my
belief in P2 and P3. The fact that there is error in P3 means I want
to continue the dialogue, so I produce output (this message – the
theoretical assumption being always that each of these control systems
has non-zero gain). That my perception and reference differ also for
P2 means I should produce output that reduces the error (I should try
to explain or illustrate my original meaning). To implement that
output is a matter for lower-level control systems (lower procol
layers, in LPT).

I perceive that you perceive my description of the nine active
control systems as “an attempt to find a statistical
generalization”, I now attempt to produce output that will lead
you toward a perception that I have been describing a process model
described in formal language, not a set of statistical observations
(this message, done both by self-referential example and by
explanation – two lower-level protocols). Your response, if any, will
affect my perceptions, particularly perceptions denoted in the theory
by Martin(Bill(P3)) and Martin(P2).

Since I have a reference that Martin(P3) and Martin(Bill(P3))
both be false (-1), if the result is Martin(Bill(P3)) → +1 (i.e.
if I perceive that you don’t want to continue the dialogue), there
will be error in my control system for the level of
Martin(Bill(P3)).

At this point, I would perceive that the dialogue had become
non-collaborative (I would be wanting to continue, but you wouldn’t),
and the GPG as drawn would cease to apply. But the control systems
would still be operating, unchanged.

The (partial state) would be that because Martin(Bill(P3)) = (+1)
and I had a reference for it to be false (-1), and (by assumption) my
gain for controlling that perception would be non-zero, I would act to
try to alter my perception Martin(Bill(P3)).

An omniscient observer might, at this point know that Bill(P3)
was factually true (you did want to abort the conversation) or that it
was false (you want to continue, but perhaps not just yet). Which of
those happened to be the case would determine whether we were in
conflict, and whether you would resist actions on my part that would
be intended to get you to continue the dialogue.

I haven’t mentioned the states of the recursions on P2, but I
should, because they affect the probable state of our individual
perceptions of the value of beliefs in P3.

For example, if I believe P2 < +1 (i.e. I don’t fully believe
you made a correct interetation, though I may not actually disbelieve
it), I probably will want to continue if I think that I have a way of
acting to increase the magnitude of my belief in P2 (i.e. get you to
come to a better understanding of what I mean, or to learn better what
your understanding actually is).

Suppose you actually believe P2 (i.e. Bill(P2) = +1; you
think you understood properly), and that is why you believe P3 (that
it’s not worth continuing), whereas Martin(Bill(P2)) < +1, then we
have a misunderstanding of a different kind, a kind that often occurs
in e-mail discussions like these.

I then would have to persuade you that your understanding of my
intent isn’t what I meant. Formally, Martin(Bill(P2)) < +1,
Bill(P2) = +1, Martin(Bill(Martin(P2))) = +1 (i.e.I perceive that you
perceive – wrongly – that I am satisfied with your interpretation).
What I do hinges on the value of this last recursion (whethe I
perceive you to believe that I am satisfied, when in fact I am
not).

I could continue for a very long, even more convoluted and boring
analysis of this situation, but I hope that what I have written so far
is enoguh to show you that " that is again an attempt to find a
statistical generalization that covers as much data as possible, not
to explain the observations in terms of an underlying model" is
wrong, and I am in fact trying to explain (or rather predict) the
observations from an underlying model.

This sort of analysis is, in my view, a control process analysis,
not an attempt to find a statistical generalization. It is a
generalization, for sure, in that the claim is that control of the
recursive beliefs about these three propositions applies in the same
way at all protocol levels.

I can see that it’s an application of PCT
to provide some theoretical infrastructure for the empirical approach.
I’m all for it, as long as I don’t have to do thattoo.

It was that, initially. Now, I think it’s an application of PCT,
the result of which forms a diagram like the GPG that describes what
most probably happens when the dialogue remains collaborative (the
references for all nine beliefs remain at +1).

Martin

[From Bill Powers (23005.05.21.1416 MDT)]

OK, but that is again an attempt
to find a statistical generalization that covers as much data as
possible, not to explain the observations in terms of an underlying
model.

I think you misunderstood much of my last message, because you couldn’t
make that statement if you had understood it as I intended.

I have goals (references) that I
should believe P1, P2, and P3. Therefore there is error in my control
systems for the values of my belief in P2 and P3.

I don’t follow the “therefore.” Where does that come
from?

The fact that there is error in
P3 means I want to continue the dialogue, so I produce output (this
message – the theoretical assumption being always that each of these
control systems has non-zero gain).

“So I produce output,” you say – why? What is the control
organization you’re talking about? Wouldn’t it be helpful to diagram the
control systems to explain what is going on here?

That my perception and
reference differ also for P2 means I should produce output that reduces
the error (I should try to explain or illustrate my original meaning). To
implement that output is a matter for lower-level control systems (lower
procol layers, in LPT).

My memory being what it is, I like to have the meanings of symbols like
these in front of me. You said

… we started with an
invented formal notation based on three statements of fact (“real
world” conditions unknowable to the participants). The three
statements were:

P1: The recipient has made an interpretation of the primal message

P2: Whenever P1 is true, then the recipient’s interpretation will be
adequate

P3: It is not worth continuing to try to reach the state (P1 &
P2).

I really don’t follow this. P2 contains the clause “Whenever P1 is
true.” We should be able to substitute the meaning of P1 for the
statement “P1 is true” in proposition P2, so it would
read
P2: Whenever the recipient has made an interpretation of the primal
message
then the recipient’s interpretation will be adequate.

[Italics are the substituted value of P1 when P1 is true.]

Is that really what you mean? That any interpretation whatsoever is
necessarily adequate? Could you possibly have meant this
instead?

P1. The recipient has made an interpretation of the primal message.

P2. The recipient’s interpretation is adequate.

Then P3 makes more sense, at least in that the expression (P1 & P2)
expresses a logical condition in which the recipient has made an
interpretation AND the interpretation is adequate.

But then proposition P3 stumps me altogether. “It is not worth
continuing to try to reach the state (P1 & P2).” Why isn’t it
worth continuing to try to reach the state where the recipient has made
an adequate interpretation? Why would you want to continue the dialogue
if it’s not worth doing? You say these are “…‘real world’
conditions unknowable to the participants”. But how could being
worthwhile be a real world condition?

Obviously I simply don’t grasp what you mean here. That being the case,
trying to follow the details of your argument is rather futile.

Best,

Bill P.

Re: Communmicative PCT (was Quantum
Thoughts)
[Martin Taylor 2005.05.21.18.21]

For the record, Bill’s questions have clarified for me an aspect
of LPT theory that has always been a little obscure. The clarification
came as I was writing this message, and I’m leaving the intermediate
stages in, unedited, so you can see (perhaps) the process.

[From Bill Powers (23005.05.21.1416
MDT)]

OK, but that is again an attempt to find
a statistical generalization that covers as much data as possible, not
to explain the observations in terms of an underlying model.

I think you misunderstood much of my last message, because you
couldn’t make that statement if you had understood it as I
intended.

I have goals (references) that I should
believe P1, P2, and P3. Therefore there is error in my control systems
for the values of my belief in P2 and P3.

I don’t follow the “therefore.”
Where does that come from?

Standard control theory. The perception differs from the
reference value in both the Martin(P2) and the Martin(P3) control
systems. Therefore there is error in those control systems.

The fact that there is error in P3 means
I want to continue the dialogue, so I produce output (this message –
the theoretical assumption being always that each of these control
systems has non-zero gain).

“So I produce output,” you say – why? What is the control
organization you’re talking about? Wouldn’t it be helpful to diagram
the control systems to explain what is going on here?

OK, I’ll try, for Martin’s control of P3. This expresses
Martin(P3) < 0, with the general assumptions of LPT for cooperative
dialogue, that the reference value is +1 and the output gain is >
0.

                      > refeernce value for Martin(P3) (+1)

           > (means I want to be in a position that

the

           > dialogue is successfully completed)

           V

 ---------comp-----------

Error > 0

Actual value of

Martin(P3) < 0

           output function
             gain > 0

perceptual |

input
function |

Martin’s |
envirnomental |
boundary

==============^=======================V==============

Things intended

Things written

to be read by Bill

by Bill

  <----Bill's

control----

systems

That my perception and reference
differ also for P2 means I should produce output that reduces the
error (I should try to explain or illustrate my original meaning). To
implement that output is a matter for lower-level control systems
(lower procol layers, in LPT).

My memory being what it is, I like to have the meanings of symbols
like these in front of me. You said

… we started with an invented
formal notation based on three statements of fact (“real world”
conditions unknowable to the participants). The three statements
were:

P1: The recipient has made an interpretation of the primal message

P2: Whenever P1 is true, then the recipient’s interpretation will be
adequate

P3: It is not worth continuing to try to reach the state (P1 &
P2).

I really don’t follow this. P2 contains the clause “Whenever P1
is true.” We should be able to substitute the meaning of P1 for
the statement “P1 is true” in proposition P2, so it would
read
P2: Whenever the recipient has made an interpretation of the primal
message
then the recipient’s interpretation will be adequate.

[Italics are the substituted value of P1 when P1 is true.]
Is that really what you mean? That any
interpretation whatsoever is necessarily adequate? Could you possibly
have meant this instead?

P1. The recipient has made an interpretation of the primal
message.
P2. The recipient’s interpretation is
adequate.

We (Dave Waugh and I) went back and forth on that, as I remember.
The reworded P2 makes no sense if P1 is believed false (the belief in
P1 has a value below zero), which is why we used the conditional form.
It might have been better to use “If”, but that has a
“crisp logic” connotation, and we are dealing with
real-valued variables. I guess one might substitute “To the
degree that” for “Whenever”, but that’s also
potentially misleading. We went with “Whenever”.

Then P3 makes more sense, at least in
that the expression (P1 & P2) expresses a logical condition in
which the recipient has made an interpretation AND the interpretation
is adequate.

Yes. At least you got through the semantic problem with P2 to get
at the intention.

But then proposition P3 stumps me
altogether. “It is not worth continuing to try to reach the state
(P1 & P2).” Why isn’t it worth continuing to try to reach the
state where the recipient has made an adequate
interpretation?

Because you are there already. So there’s no point in continuing to
try to get there.

Why would you want to continue the
dialogue if it’s not worth doing?

You don’t. P3 is phrased in that negative way so as to have a +1
reference value for all three propositions.

P3 can have a value +1 if one is frustrated and perceives that
the dialogue is going nowhere (as happens from time to time on
CSGnet). Perceiving P3 to be true stops one’s contribution to the
dialogue, and therefore stops the dialogue. Perceiving the other to
perceive P3 to be true doesn’t stop it, but one’s P3 control system
produces output that tries to get the other to perceive P3 to be
false, because if the other perceives P3 to be true, they will stop
contributing to the dialogue, breaking the feedback loop. The dialogue
will actually stop, but one doesn’t want it to stop, so one acts to
restart it.

You say these are “…‘real
world’ conditions unknowable to the participants”. But how could
being worthwhile be a real world condition?

Good point. P3 exists only as a belief in one participant about a
state of that participant, whereas P1 and P2 exist in one participant
about a state in the other. [Note added afterwards: This was the
observation that clarified for me the sticky point].

Obviously I simply don’t grasp what you
mean here. That being the case, trying to follow the details of your
argument is rather futile.

Equally obviously, I’ve not been as precise as I had thought I
was being.

On thinking about it, I think it might be better to split P3, and
consider two “stop” conditions. One is implicit in P1 and P2
being believed true (along with their recursions). If those are true,
they are equal to their reference values, and hence their output will
stop. The outputs from the P1 and P2 control systems at different
levels of recursion contribute the content of the primal message. If
all of O(P1 & P2), O(R(P1 & P2)), O(R(O(P1&P2))) and the
corresponding recipient beliefs are +1, then the dialogue will have
ended without any explicit decision to end it.

I have always considered this “success” condition to be
an input to the P3 perceptual function, "OR"ed with the
other condition (“frustration and failure”, in effect) that
leads to the end of the dialogue that passes that primal message. But
the signals that contribute to the failure condition are derived from
different sources unrelated to the content of the primal message, and
it might be better to think of them as creating another perception
whose control output influences the gain of the output functions for
controlling the belief in P1 and P2.

I don’t think there’s any need to change the three propositions
(apart from finding a way to avoid the “Whenever” ambiguity
in P2), but the “stop” state for the passing of one primal
message is changed. The dialogue stops “naturally” when all
the P1 and P2 belief recursions are at their reference levels, OR when
P3, whose belief value does not depend on the state of the passing of
the primal message, is true. This last is modelled by causing the
output of the O(P3) control system to modulate the gain of the O(P1)
and O(P2) systems.

Subjectively, this feels right to me. Persistent failure to get a
message across often leads to a state in which I may still have a
reference to get it across, but no longer feel I have the power to do
so, so I stop trying.

I think your comments have advanced LPT theory. Thank you
for that.

Martin

[From Bill Powers (2005.05.22.0954 MDT)]

Martin Taylor 2005.05.21.18.21–

The fact that there is error in
P3 means I want to continue the dialogue, so I produce output (this
message – the theoretical assumption being always that each of these
control systems has non-zero
gain).

So I understand from this that the perception for condition 3 is “It
is worthwhile to continue the dialog,” while the reference is
“it is not worthwhile to continue dialogue”, and the difference
is why the dialog continues.
This doesn’t look to me like the best way to express the situation. It
seems to me that the error to be corrected by the dialog is simply that
you want the other person to show signs of understanding (reference),
while that person does not yet show such signs (perception). The error
maintains the reference for continuing to look for ways to clarify or
explain.
I have some suggestions, but I have to start with some further analysis
of P1 and P2.
The perceptual input function producing P3 as presented so far is AND
(&) with the input arguments being P1, P2. The perceptual signal is
the value of this function, which is 1 if and only if p1 and p2 are both
true (p1 & p2 = 1). The reference signal is 1, which has the
meaning “p1 and p2 are both true” but of course is not really
specific to the two inputs. The error signal is 1 if the perceptual
signal has the same value as the perceptual signal, so the comparator
function for this logical system is e = (**not-**p &
r) or (p & *not-***r), which has the
value of 0 (no error) when p and r have the same value, whether 1 or 0.
That’s the exclusive or.
Actually, the perceptual signal as I revised it isn’t correct. I would
infer that you have formed an interpretation of my output if you say
anything that I perceive as describing any interpretation. So whether the
interpretation is adequate or not, I still perceive that you have at
least made an interpretation. The P1 clause is
therefore unnecessary (you could hardly make an inadequate interpretation
if you had not made an interpretation). Only the P2 clause
matters.

The P2 clause is necessary because I am trying to communicate a meaning,
and I judge the adequacy of your (inferred) interpretation by its ability
to evoke a meaning in me that matches the meaning I want to perceive. I
may want you to understand what I mean by saying “today is
Tuesday,” so if you respond by saying “I see that you’re saying
that today is the day after Monday,” the meaning I get from the
paraphrase may be the meaning I intend, so I am satisfied (zero error).
On the other hand, if my meaning is “You should go to the polls and
vote” (Tuesday is the irrational traditional voting day in the US),
I will not consider your response adequate and will continue to seek a
way to get you to show that you realize that this is Election Day.
Perhaps your just-prior communication was “I plan to vote tomorrow,
do you?”

So I think your LP scheme needs to include the distinction between a
communication and the meaning of a communication. I’m sure this
distinction is made somethere within the theory, but it should be made
clear when saying what the term “adequate interpretation”
means.

The difference between your communication and my intentions for your
communication is not in the words used to communicate (that statement is
surely part of LPT) but in the meanings of the words. It is the
difference between the meanings of the words as nonverbal perceptions
(Benjamin Whorf notwithstanding). If I want you to mail me the
robin’s-egg-blue paint sample and you ask if the light blue one will do,
I will say OK, but I won’t say OK if you ask me if I’ll accept the violet
one. Violet has too much red in it. That is not because there is too much
of the word “red” in the word “violet”, but because
there is too much red in the perception of violet that the word
“violet” points to in my own remembered experiences. The error
is the usual kind between a color perception and a color reference
signal, the main difference being that the perception involved is being
generated by a verbal communication rather than by the retina in the
usual way. That generation process is, of course, a very useful capacity
of the recipient’s brain without which there would be no language, and
the meanings that are evoked come from the recipient’s stored
experiences, not the speaker’s.

That’s enough for one bite.

Best,

Bill P.

Re: Communmicative PCT (was Quantum
Thoughts)
Martin Taylor 2005.05.24.14.55]

[From Bill Powers (2005.05.22.0954
MDT)]

Martin Taylor 2005.05.21.18.21–

The fact that there is error in P3 means
I want to continue the dialogue, so I produce output (this message –
the theoretical assumption being always that each of these control
systems has non-zero gain).

So I understand from this that the perception for condition 3 is
“It is worthwhile to continue the dialog,” while the
reference is “it is not worthwhile to continue dialogue”,
and the difference is why the dialog continues.

This doesn’t look to me like the best way to express the situation. It
seems to me that the error to be corrected by the dialog is simply
that you want the other person to show signs of understanding
(reference), while that person does not yet show such signs
(perception). The error maintains the reference for continuing to look
for ways to clarify or explain.

So far, so good. The insight your earlier message gave me was
that the two reasons for terminating the dialogue should be separated,
whereas hitherto, the behavioural result was what I had been looking
at, and thinking of that result as being the consqeunce of one control
mechanism. I was not using a fully PCT-based approach.

Clearly (now:-), one reason the dialogue stops is that it has
achieved its desired result. Both parties are satisfied that their
goals have been met: the Originator perceives that the Recipient’s
state has been moved to its reference value, and the Recipient
perceives that the Originator is satisfied, which (in a collaborative
dialogue) is the Recipient’s main reference condition.

The other reason the dialogue stops is that the gain of the P1
and/or P2 control system in either party goes to zero. Even though
there may still be error, in that the Recipient seems not to have made
an interpretation, or if one has been made, it seems to be wrong, one
of the parties is simply tired of the effort to reduce the error. They
give up. The imputs to this control system are not related to the
values of P1 and P2 beliefs, but they may be related to the
derivatives of those values and of some function of the time spent on
the dialogue, or there may be a conflict issue, as when one of the
parties has to do something else more important.

I have some suggestions, but I have to
start with some further analysis of P1 and P2.
The perceptual input function producing
P3 as presented so far is AND (&) with the input arguments being
P1, P2.

(As of last Saturday, that changed. The “frustration”
inputs now are the only ones – in my new vision of the theory).

The perceptual signal is the value
of this function, which is 1 if and only if p1 and p2 are both true
(p1 & p2 = 1). The reference signal is 1, which has the
meaning “p1 and p2 are both true” but of course is not
really specific to the two inputs. The error signal is 1 if the
perceptual signal has the same value as the >reference< signal,
so the comparator function for this logical system is e =
(**not-**p & r) or (p *& not-***r),
which has the value of 0 (no error) when p and r have the same value,
whether 1 or 0. That’s the exclusive or.

Remember that the values are real, so p1 could be 0.523, giving an
error of 0.477. By assumption, in a collaborative dialogue, the
reference values are always 1.0.

Actually, the perceptual signal as I revised it isn’t correct. I would
infer that you have formed an interpretation of my output if you say
anything that I perceive as describing any interpretation. So whether
the interpretation is adequate or not, I still perceive that you have
at least made an interpretation. The P1 clause is therefore
unnecessary (you could hardly make an inadequate interpretation if you
had not made an interpretation). Only the P2 clause
matters.

Not true. The originator’s behaviour is different if O(P1) <
0, or eve perhaps if it is slightly positive. If the Recipient seems
to have made some sort of an interpretation, then the Originator’s job
is to find out how that interpretation differs from what the
Originator intended. If not, the Originator’s job is to try to find
out what kind of language would allow the Recipient to start making an
interpretation.

The P2 clause is necessary because I am
trying to communicate a meaning, and I judge the adequacy of your
(inferred) interpretation by its ability to evoke a meaning in me that
matches the meaning I want to perceive.

Yes, provided that you take “meaning” in a rather
general sense. In LPT, “meaning” doesn’t mean
“understanding what I meant” or “transferring an idea
from one brain to another.” It means “achieving in the
recipient the desired state”, which might mean seeing the
Recipient wordlessly open a window.

So I think your LP scheme needs to
include the distinction between a communication and the meaning of a
communication. I’m sure this distinction is made somethere within the
theory, but it should be made clear when saying what the term
“adequate interpretation” means.

It is not only made clear, it is central to the theory. A
communication is a means whereby a meaning is produced. It’s like the
difference between a sound from a loudspeaker and the wire that con
transmit electrical current.

“Meaning” and “adequate interpretation” are
not exactly the same thing, though they are close. An “adequate
interpretation” is something only the Originator can determine,
and he can determine it only by observing whether the Recipient’s
state is close enough to his reference for it that he no longer wishes
to alter the Recipient’s state. The “meaning” is the actual
change of state, which might possibly be perceptible by the Recipient.
The “intended meaning” is the change of Recipient State
intended by the Originator – in other words, it is the difference
between the Originator’s perception of the Recipient’s state at the
initiation of the dialogue and its reference value.

The difference between your communication
and my intentions for your communication is not in the words used to
communicate (that statement is surely part of LPT) but in the meanings
of the words.

No, it’s more than that. For one thing, no words need be involved.
Iago’s message to Othello (not in a collaborative dialogue, to be
sure) had a very clear meaning, which was successfully transmitted. It
consisted of a dropped handkerchief.

Even in messages transmitted through the medium of words, it is
often very hard to determine the meaning of the message by reference
to the dictionary meanings of the words. The message may be (and often
is) in the connotations, in context of the previous interactions
between the parties.

It is the difference between the
meanings of the words as nonverbal perceptions (Benjamin Whorf
notwithstanding). If I want you to mail me the robin’s-egg-blue paint
sample and you ask if the light blue one will do, I will say OK, but I
won’t say OK if you ask me if I’ll accept the violet one.

Fair enough. You, as Originator, had a message to send. I, as
Recipient, had to determine what that message was. From your point of
view, an “adequate interpretation” was bounded somewhere
between my sending you a light blue one and a violet one. In the GPG
you sent a primary message asking for the sample. I perceived that I
did not have a strong belief in P2, and therefore sent a message
asking for clarifiaction (which allowed you to perceive I did not have
a strong belief in P2). If I had asked “is light blue OK”,
your would have increased your belief in P2, and by responding
“OK”, you enhanced my belief not only in P2, but my belief
that you believed I believed P2. The completion of the dialogue would
be your receipt of an acceptable paint sample.

If I had sent you a violet sample, without any verbal comment,
the dialogue would be incomplete, since I had shown you that you
should believe P2 to be false, leaving it up to you to do what you
could to bring me toward a correct interpretation.

Violet has too much red in it. That
is not because there is too much of the word “red” in the
word “violet”, but because there is too much red in the
perception of violet that the word “violet” points to in my
own remembered experiences. The error is the usual kind between a
color perception and a color reference signal, the main difference
being that the perception involved is being generated by a verbal
communication rather than by the retina in the usual way.

It really doesn’t matter to the theory what I do that allows you to
perceive the state of my understanding. Whether it’s words, gestures,
or more comples actions (euch as, perhaps, starting a multi-year
campaign for a Senate seat), it’s you observation of me that allows
you to compare SOMETHING to your reference for what that something
“should” be. Even if I say exactly the right things, if I do
something that differs from your reference for what the message should
have achieved, you won’t believe P2. Of course, you may believe that I
understand what you want, and am being just bloody-minded or hostile,
but that gets us into the realm of non-collaborative dialogue.

That generation process is, of course, a
very useful capacity of the recipient’s brain without which there
would be no language, and the meanings that are evoked come from the
recipient’s stored experiences, not the speaker’s.

Those experiences, on different time scales, are an important
part of LPT that I haven’t started to discuss here. Until the
interplay of controlling the levels of belief in the three
propositions is clear, the means whereby those beliefs change are more
likely than not to complicate the discussion.

That’s enough for one bite.

It’s all rather a mouthful.

I think you can drop consideration of P3 for the time being, and
consider what happens in a collaborative dialogue for different states
of the (now) 6 control systems in each partner. Some parts of that
12-D space are unlikely to occur, and hard to justify rationally,
though they are possible. For example, it is possible that the
Recipient has made a correct interpretation, but thinks he has not
made any interpretation (R(P1) < 0, R(P2) = 1), but it’s hardly
ever going to happen, so we ignore it, and blank off that whole
hypervolume. Much more likely is R(P2 = 1), O(P2) = -1. The Recipient
is sure he has understood, while the originator is quite clear he
hasn’t. That happens a lot on CSGnet:-)

Martin

[From Bill Powers (2005.05.24.2046 MDT)]

Martin Taylor 2005.05.24.14.55 --

Clearly (now:-), one reason the dialogue stops is that it has achieved its desired result. Both parties are satisfied that their goals have been met: the Originator perceives that the Recipient's state has been moved to its reference value, and the Recipient perceives that the Originator is satisfied, which (in a collaborative dialogue) is the Recipient's main reference condition.

I think we're in agreement to here.

The other reason the dialogue stops is that the gain of the P1 and/or P2 control system in either party goes to zero.

This isn't the only way to make it stop, and it has some disadvantages (possibly). If the output gain is brought to zero while an error still exists, the error will remain; only the willingness to do something about it goes away. Also the desire to achieve understanding (the reference signal) is still there, unsatisfied, which isn't a satisfactory ending for higher-level systems, either.

I think a better way to handle this is with a higher-order system that sets the reference level (or if you like, the gain) for the lower communication system in the first place. Perhaps this is where your idea of "worth continuing" belongs. If this higher system judges the dialog to be worth continuing, it maintains a nonzero reference level for getting the other party to understand (and something symmetrical happens in the other party, I presume). If the higher order system judges that progress is being made, or that there is some other benefit being realized, the reference or gain is maintained high. But if the process bogs down or starts going in circles, the higher system sees this as unsatisfactory and looks for some other way to achieve satisfaction -- setting this reference signal or gain to zero and selecting other subsystems to use instead. I see that I'm assuming that the higher system has a goal of finding pursuits that are judged worth continuing, and while error exists does so by scanning over possible lower-order systems to find one that satisfies the goal.

I don't know where this fits in with my definitions of levels; it could be like "fine structure" within the program level. Fitting with my definitions shouldn't be a constraint on LPT, of course.

Even though there may still be error, in that the Recipient seems not to have made an interpretation, or if one has been made, it seems to be wrong, one of the parties is simply tired of the effort to reduce the error. They give up.

Right. That's a higher-level decision, isn't it? Control systems don't spontaneously give up; they are told to quit, or are turned off.

I have some suggestions, but I have to start with some further analysis of P1 and P2.
The P1 clause is therefore unnecessary (you could hardly make an inadequate interpretation if you had not made an interpretation). Only the P2 clause matters.

Not true. The originator's behaviour is different if O(P1) < 0, or even perhaps if it is slightly positive. If the Recipient seems to have made some sort of an interpretation, then the Originator's job is to find out how that interpretation differs from what the Originator intended. If not, the Originator's job is to try to find out what kind of language would allow the Recipient to start making an interpretation.

Well, this adds more features to the model. Now we have the reference condition "Recipient is making some sort of interpretation," with the means of controlling that variable being finding language that results in meeting that goal.

The P2 clause is necessary because I am trying to communicate a meaning, and I judge the adequacy of your (inferred) interpretation by its ability to evoke a meaning in me that matches the meaning I want to perceive.

Yes, provided that you take "meaning" in a rather general sense. In LPT, "meaning" doesn't mean "understanding what I meant" or "transferring an idea from one brain to another." It means "achieving in the recipient the desired state", which might mean seeing the Recipient wordlessly open a window.

Yes, and this would imply that the recipient has converted your words into reference signals for perceptions that coincide with what you wanted to communicate -- an open window, or a cooler (or warmer) room, or a better view.

So I think your LP scheme needs to include the distinction between a communication and the meaning of a communication. I'm sure this distinction is made somethere within the theory, but it should be made clear when saying what the term "adequate interpretation" means.

It is not only made clear, it is central to the theory. A communication is a means whereby a meaning is produced. It's like the difference between a sound from a loudspeaker and the wire that can transmit electrical current.

OK, good.

"Meaning" and "adequate interpretation" are not exactly the same thing, though they are close. An "adequate interpretation" is something only the Originator can determine, and he can determine it only by observing whether the Recipient's state is close enough to his reference for it that he no longer wishes to alter the Recipient's state. The "meaning" is the actual change of state, which might possibly be perceptible by the Recipient. The "intended meaning" is the change of Recipient State intended by the Originator -- in other words, it is the difference between the Originator's perception of the Recipient's state at the initiation of the dialogue and its reference value.

If you use "meaning" for "actual change of state," what do you use to indicate the nonverbal perceptions that the originator experiences as a result of receiving the communication from the recipient? I'm not using "meaning" in the sense of "significance" or "implications." Are you?

The difference between your communication and my intentions for your communication is not in the words used to communicate (that statement is surely part of LPT) but in the meanings of the words.

No, it's more than that. For one thing, no words need be involved. Iago's message to Othello (not in a collaborative dialogue, to be sure) had a very clear meaning, which was successfully transmitted. It consisted of a dropped handkerchief.

Don't quibble. The handkerchief was a symbolic communication. It's meaning was the perception of betrayal that Othello experienced (imagined, but still a perception).

Even in messages transmitted through the medium of words, it is often very hard to determine the meaning of the message by reference to the dictionary meanings of the words. The message may be (and often is) in the connotations, in context of the previous interactions between the parties.

I'd wrap that up by saying that meaning consists of the perceptions evoked within the recipient upon recieving (or failing to receive) a communication.

It is the difference between the meanings of the words as nonverbal perceptions (Benjamin Whorf notwithstanding). If I want you to mail me the robin's-egg-blue paint sample and you ask if the light blue one will do, I will say OK, but I won't say OK if you ask me if I'll accept the violet one.

Fair enough. You, as Originator, had a message to send. I, as Recipient, had to determine what that message was. From your point of view, an "adequate interpretation" was bounded somewhere between my sending you a light blue one and a violet one. In the GPG you sent a primary message asking for the sample. I perceived that I did not have a strong belief in P2, and therefore sent a message asking for clarifiaction (which allowed you to perceive I did not have a strong belief in P2). If I had asked "is light blue OK", your would have increased your belief in P2, and by responding "OK", you enhanced my belief not only in P2, but my belief that you believed I believed P2. The completion of the dialogue would be your receipt of an acceptable paint sample.

Agreed. It comes out very nicely.

If I had sent you a violet sample, without any verbal comment, the dialogue would be incomplete, since I had shown you that you should believe P2 to be false, leaving it up to you to do what you could to bring me toward a correct interpretation.

Violet has too much red in it. That is not because there is too much of the word "red" in the word "violet", but because there is too much red in the perception of violet that the word "violet" points to in my own remembered experiences. The error is the usual kind between a color perception and a color reference signal, the main difference being that the perception involved is being generated by a verbal communication rather than by the retina in the usual way.

It really doesn't matter to the theory what I do that allows you to perceive the state of my understanding. Whether it's words, gestures, or more comples actions (euch as, perhaps, starting a multi-year campaign for a Senate seat), it's you observation of me that allows you to compare SOMETHING to your reference for what that something "should" be.

All right, strike "verbal" and retain "communication."

Even if I say exactly the right things, if I do something that differs from your reference for what the message should have achieved, you won't believe P2. Of course, you may believe that I understand what you want, and am being just bloody-minded or hostile, but that gets us into the realm of non-collaborative dialogue.

Yes, it's my perception of the result that I compare with my intention. If I just want you to act in a certain way that I can see, or that affects me, I will be satisfied when you do it. If I also want you to understand what I meant, I will also demand that you say things that evoke the right meanings in me. This is wanting you to do the right thing, but for the right reasons.

I think you can drop consideration of P3 for the time being, and consider what happens in a collaborative dialogue for different states of the (now) 6 control systems in each partner. Some parts of that 12-D space are unlikely to occur, and hard to justify rationally, though they are possible. For example, it is possible that the Recipient has made a correct interpretation, but thinks he has not made any interpretation (R(P1) < 0, R(P2) = 1), but it's hardly ever going to happen, so we ignore it, and blank off that whole hypervolume. Much more likely is R(P2 = 1), O(P2) = -1. The Recipient is sure he has understood, while the originator is quite clear he hasn't. That happens a lot on CSGnet:-)

Amen.

I think I will avoid getting further into this, not because it isn't interesting but because I'm juggling too many things again, as well as being driven to utter frustration by trying to learn opaque graphics programming languages, by which I mean not opaque graphics but opaque languages. I started on openGL and managed to get some programs running, but they were too slow so I got a book on DirectX, which proves to have something even the author calls a "steep learning curve." The ratio of things you have to do to make the language happy to things you have to do to actually create a result is enormous. My late friend Walter Weller called all that getting-ready stuff "foreplay." It goes on long after I'm wanting to get to the point.

Best,

Bill P.

[Martin Taylor 2005.05.27.22.06]

[From Bill Powers (2005.05.24.2046 MDT)]

Martin Taylor 2005.05.24.14.55 --

The other reason the dialogue stops is that the gain of the P1 and/or P2 control system in either party goes to zero.

This isn't the only way to make it stop, and it has some disadvantages (possibly). If the output gain is brought to zero while an error still exists, the error will remain; only the willingness to do something about it goes away. Also the desire to achieve understanding (the reference signal) is still there, unsatisfied, which isn't a satisfactory ending for higher-level systems, either.

But that's exactly the point! The desire to get the other to understand hasn't gone away. It is frustrated. There are several possible reasons: conflict with some other control system competing for resources (e.g. a prior appointment terminates a conversation), the recipient has ceased to be collaborative (no longer has a reference to be perceived by the originator as wanting to understand), the originator can not find a mechanism to advance the recipient's understanding...

I think a better way to handle this is with a higher-order system that sets the reference level (or if you like, the gain) for the lower communication system in the first place.

That often happens, in LPT. It's a given that the need for the recipient to "understand" a primal message may go away if the circumstances change. A typical reason for such a change is that a higher-level primal message to which this one contributes has been satisfactorily understood. That's anothe way of saying, in PCT language, that the reference level for getting the primal message understood has gone to zero.

Even though there may still be error, in that the Recipient seems not to have made an interpretation, or if one has been made, it seems to be wrong, one of the parties is simply tired of the effort to reduce the error. They give up.

Right. That's a higher-level decision, isn't it? Control systems don't spontaneously give up; they are told to quit, or are turned off.

"Are turned off" is, to me, the same as "their gain is set to zero". Do you mean something else?

I have some suggestions, but I have to start with some further analysis of P1 and P2.
The P1 clause is therefore unnecessary (you could hardly make an inadequate interpretation if you had not made an interpretation). Only the P2 clause matters.

Not true. The originator's behaviour is different if O(P1) < 0, or even perhaps if it is slightly positive. If the Recipient seems to have made some sort of an interpretation, then the Originator's job is to find out how that interpretation differs from what the Originator intended. If not, the Originator's job is to try to find out what kind of language would allow the Recipient to start making an interpretation.

Well, this adds more features to the model. Now we have the reference condition "Recipient is making some sort of interpretation," with the means of controlling that variable being finding language that results in meeting that goal.

It's no different than any other higher-level control system in one-person PCT. The way that a perception is influenced by the output of its control system can be very complicated, in the passsage through many lower levels of control. I don't think any new feature has been added either to the LPT model orto its PCT restatement.

I certainly used language that applies to many levels at once, but it's the same kind of statement as "If I want to be able to make this contraption work, I have to find out why it doesn't, and fix it." That doesn't seem to require any new feature in PCT, or if it does, it's the same feature.

"Meaning" and "adequate interpretation" are not exactly the same thing, though they are close. An "adequate interpretation" is something only the Originator can determine, and he can determine it only by observing whether the Recipient's state is close enough to his reference for it that he no longer wishes to alter the Recipient's state. The "meaning" is the actual change of state, which might possibly be perceptible by the Recipient. The "intended meaning" is the change of Recipient State intended by the Originator -- in other words, it is the difference between the Originator's perception of the Recipient's state at the initiation of the dialogue and its reference value.

If you use "meaning" for "actual change of state," what do you use to indicate the nonverbal perceptions that the originator experiences as a result of receiving the communication from the recipient? I'm not using "meaning" in the sense of "significance" or "implications." Are you?

No, and I'm not using "meaning" as meaning the same for originator and recipient, either. For the originator, the meaning of the primal message is the intended change of the recipient's state. For the recipient, it's the actual change of state. The originator in face to face communication probably would be able to use nonverbal observations as part of the input to the perception of the recipient's actual change of state, and would use them to determine the error still left in trying to get across the intended meaning.

There's nothing to say that the recipient won't pick up messages unintended by the originator. They may be non-verbal, or in intonation, or in word choice... And there's no reason to assert that the originator will perceive that the recipient has picked up those messages. But they aren't part of collaborative dialogue. LPT incorporates them, but the General Protocol Grammar doesn't.

The difference between your communication and my intentions for your communication is not in the words used to communicate (that statement is surely part of LPT) but in the meanings of the words.

No, it's more than that. For one thing, no words need be involved. Iago's message to Othello (not in a collaborative dialogue, to be sure) had a very clear meaning, which was successfully transmitted. It consisted of a dropped handkerchief.

Don't quibble. The handkerchief was a symbolic communication. It's meaning was the perception of betrayal that Othello experienced (imagined, but still a perception).

It's not a quibble at all. It illustrates a central point, that words are just a protocol level in the theory, just one possible way of passing a message. Dropping a handkerchief was another way -- that it was in a noncollaborative dialogue is irrelevant. A hundred years ago, a dropped handkerchief was sometimes the equivalent of the words "Hello young man, would you like to talk with me?" I don't see what your use of the term "symbolic communication" contributes. Aren't words "symbolic communications" too?

Even in messages transmitted through the medium of words, it is often very hard to determine the meaning of the message by reference to the dictionary meanings of the words. The message may be (and often is) in the connotations, in context of the previous interactions between the parties.

I'd wrap that up by saying that meaning consists of the perceptions evoked within the recipient upon recieving (or failing to receive) a communication.

They could hardly be evoked on failure to receive the communication, could they? That would mean the recipient had not perceived any event as having been caused by the originator.

I think I will avoid getting further into this, not because it isn't interesting but because I'm juggling too many things again,

Understood, but I'll be sorry, since your questions have clarified stuff for me.

as well as being driven to utter frustration by trying to learn opaque graphics programming languages, by which I mean not opaque graphics but opaque languages. I started on openGL and managed to get some programs running, but they were too slow so I got a book on DirectX, which proves to have something even the author calls a "steep learning curve."

It's expensive, but have you ever considered Labview, from national Instruments? Allan Randall says there is a much cheaper version intended for programming robots, but it has most of the features of Labview. I've forgotten its name, but if you are interested, I'll ask.

Ooops...I see you meant "opaque textual languages for programming graphics". Nevertheless, you still might be interested in "Labview Light"

The ratio of things you have to do to make the language happy to things you have to do to actually create a result is enormous. My late friend Walter Weller called all that getting-ready stuff "foreplay." It goes on long after I'm wanting to get to the point.

All so true.

Martin

[Martin Taylor 2005.05.28.20.39]

[From Bill Powers (2005.05.24.2046 MDT)]

Martin Taylor 2005.05.24.14.55 --

The other reason the dialogue stops is that the gain of the P1 and/or P2 control system in either party goes to zero.

This isn't the only way to make it stop, and it has some disadvantages (possibly). If the output gain is brought to zero while an error still exists, the error will remain; only the willingness to do something about it goes away. Also the desire to achieve understanding (the reference signal) is still there, unsatisfied, which isn't a satisfactory ending for higher-level systems, either.
...
I think I will avoid getting further into this, not because it isn't interesting but because I'm juggling too many things again, as well as being driven to utter frustration by trying to learn opaque graphics programming languages, by which I mean not opaque graphics but opaque languages.

Did you notice that this is the condition in question: Bill(P1) near +1 (little or no error), Bill(P2) positive but small (substantial error), Bill(P3) = +1 (which aborts the dialogue)?

If I take you at your word, the reason for ending the dialogue isn't that the reference for P2 has changed, but that there is conflict, and you have to choose whether to continue controlling for Bill(P2), or to control some quite different perception. Conflict is one of the states that I suggested to you as affecting the value of Person(P3).

Also, although it may be obvious, the values I ascribed to Bill(Pn) are actually the values of Martin(Bill(Pn)), based on your writings. I have no access to the values of any beliefs for which the first label is "Bill".

Martin

[From Bill Powers (2005.05.29.0030 MDT)]

Martin Taylor 2005.05.28.20.39 --

If I take you at your word, the reason for ending the dialogue isn't that the reference for P2 has changed, but that there is conflict, and you have to choose whether to continue controlling for Bill(P2), or to control some quite different perception. Conflict is one of the states that I suggested to you as affecting the value of Person(P3).

A slight conflict. My error relative to P2 is now sufficiently low that other errors demand attention. But LPT obviously works!

Best,

Bill P.