[Martin Taylor 940722 18:00]
Avery Andrews 940723.1400
Very quickly, before I go off-line for what might be a couple of weeks.
I've spent a certain amount of time over the last year or so reading
some of Martin Taylor's writings on Layered Protocol theory, & have
a few queries to try to focus my thoughts a bit (I dont' claim to have
read them thoroughly or well).
More so than most, I'll wager!
I think get the general idea of higher-level protocol nodes calling
on lower-level ones, but I can't summon up any kind of concrete example
of how this idea might apply to ordinary conversation, such, as, for
example, when my wife & I try to figure out who takes the car and who
takes a bicycle to work in the morning.
Do you have "Dialogue Analysis using Layered Protocols"?
The base notion (now; more precision than in anything you yet have seen)
is that the primal message in each protocol node is strictly analogous to
a reference signal in a PCT hierarchy. In LPT, the "signals" are not
scalars, but the concept is the same. The primal message represents a
state of the communicating partner that the originator of the message wants
to perceive. The virtual message (the output of the Protocol Node) that is
sent is the difference between the currently perceived state and the
reference state, and the "currently perceived" state is continuously
monitored (as far as possible) by the message originator.
In your case, you start from what I call a "task-level" reference, of
perceiving both persons to be happily transported to work in the morning.
You could perhaps achieve this without communicating verbally, but you
choose to discuss the matter. In LPT, the top-level primal message is
the discrepancy between you both now being at home and you both being
happily transported. There is imagined (planned?) perception involved
on both sides. You perceive yourself happily taking the bike and your
wife happily taking the car, but you don't know how she perceives the
possibilities. You have a reference perception that you DO know what
she perceives, so you have an error signal that induces output of the
kind that reduces your error. Past reorganization has created the
appropriate connections--you ask what she wants.
How do you do this "asking?" This is a matter of convention, as I discussed
a couple of weeks ago. You two, in context of a culture, have developed
ways to indicate to each other that certain actions would help in reducing
error signals. Perhaps you put your hand on the bike, while raising your
eyebrows. Perhaps you say "Should I take the bike today?" Perhaps you
toss a coin in the air. Any of these are ways to disturb her so that she
has errors in some controlled perceptions relating to you; in particular,
you presume she is controlling for a perception that you do not have
unanswered questions, or something like that. She send a virtual message
to you to let you know she is happy to take the car. She may implement
this message by pointing to the car and then to herself; she may implement
it by smiling as you ask; she may implement it by saying "Sure, that's
fine by me." All of these are lower-level protocol ways to send the same
primal message--that she perceives you to know she is happy for you to
take the bike.
I've contemplated the idea
that one node might be `semantics', which calls on 'syntax', which
calls on 'phonology/phonetics', but I'm not sure that anything like
this is intended,
Absolutely, definitively NOT. In LPT, EVERY protocol incorporates its
own syntax and lexicon, as well as (possibly) pragmatics. The concept
of "semantics" is usually unclearly distinguished from "syntax" on one
side, and "pragmatics" on the other. Within LPT, "semantics" refers to
the relationship between supporting and supported protocols. In the
above example, if your wife indicates her satisfaction with taking the
car by smiling, she is using a protocol that supports the message that
she is happy about it. She could have indicated unhappiness or indecision
by other kinds of gesture, and the relation between the different possible
gestures and the intentions they signify would be the semantics of those
two protocols. Most such relationships would be dependent on the situational
context, in exactly the same way as the perceptions in other parts of the
PCT hierarchy may incorporate low-level perceptual signals of many kinds,
from many places.
We are hoping to contribute a paper on PCT and LPT in the design of
interfaces for the special issue of IJHCS. I broke off editing it to
post this response.
I hope this is useful to you. Also, I am aware that the shortness of this
response leads me to use terminology that is open to misunderstanding. I
trust that the "old hands" will for once resolve ambiguities in the sense
that I do understand at least a little about PCT.
Martin