[From Bruce Gregory 9980508.1245 EDT)]
Rick Marken (980508.0910)
Bill Powers (980508.0128 MDT) --
> In HPCT, we treat conflict as basically a mistake. In a
> properly-organized system, a higher level of control never
> sets two effectively incompatible reference signals for
> lower-level systems at the same time... That strikes me as
> quite a different model from the one you guys are agreeing on.
Bruce Gregory (980508.0510 EDT) --
> I don't think we disagree at all. What I was trying to describe was
> a situation in which the sequence was something like "talk to this
> guy until you have to leave for next appointment".
That's a contingency (program), not a sequence. A sequence is "talk
to this guy (for x minutes) then leave for appointment".
> If the conversation is not too engrossing, this works fine.
Yes. The contingency ends up being the same as the sequence in this
case. But if the conversation ends up being engrossing, you're
right there in the conflict; control of the contingency hasn't
helped solve the conflict. (How does a conversation become
engrossing, by the way? Is engrossing a perceptual aspect of a
conversation? Do conversations really _engross_? Do pretty girls
passing by really _call forth_?
A conversation becomes engrossing when leaving it increases the error
associated with the perception you are controlling.
> If the conversation _is_ engrossing the error generated by not
> departing grows as you pass your self-appointed deadline.
This sounds like a hybrid S-R/control model to me. You seem to be
saying that the "engrossingness" of the conversation keeps you in
it (that's the S-R part). The continuation of the conversation
(do to the engrossingness) creates error in the system controlling
for meeting the deadline (that's the control part).
Yes indeed. I was too brief in my exposition. Thanks for clarifying it.
> Now a higher level system must decide what to do.
In Bill's model, the higher level system is always there, controlling
the perception of sequence ("conversation then meeting"). Higher level
systems are not "triggered" by error in lower level systems (this
is another S-R concept). The higher level system, if it exists, is
always there, making sure that the perception it wants (in this
case the perception of the sequence "conversation then meeting")
is occurring.
Again thanks for the clarification. This is what I intended to say.
> Is the error associated with missing the next appointment greater
> than the error associated with breaking off the conversation?
Now you are describing a higher level control system that is
controlling a perception of the relative size of the error in
two lower level systems. I don't think this sort of design will
work. Anyway, it is certainly a very different model than the
one Bill described.
O.k. Now we have a real difference. I'd appreciate understanding why this is
so. Isn't this the principle on which the reorganization system works?
> What I was saying was that _ultimately_ you will "decide" in a way
> that minimizes total system error.
But this says very little other than that negative feedback control
systems act to reduce error; total error in a collection of control
systems is always minimized. The problem is that, when there is
conflict in a collection of control systems, the minimum to which
total error can be reduced is _much_ larger than it is when there is
no conflict (this can easily be seen by introducing a conflict into
my spreadsheet model of a hierarchy of control systems).
Yes, and exactly how is this a problem?
I think what may be missing from your understanding of the PCT
model of conflict is that conflict cannot be solved at the level
of the conflict itself.
Quite the contrary, this is very clear to me.
Conflicts can only be solved by some
form of reorganization -- reorganization that changes the _way_
the _goals_ for the conflicting systems are set. If you already
have a higher level system in place (like a sequence control system)
then there will be no conflict (like the conversation vs meeting
conflict); you will just have a disturbance to a higher level
perception (like "having a satisfying social life") that can be
solved by selecting an appropriate reference for the perception of
a sequence (have conversation then attend meeting, or vice versa).
You always have a higher order system "in place", or so it seems to me. If I
understand you correctly, a conflict by definition cannot be solved except
by reorganization. This is not true, of course, for conflicts between
hierarchical control systems, since there one system can reset the reference
level that is producing the conflict in order to realize a higher level
goal. The Israelis and the Palestinians can want a village to be in their
domains. One can concede in exchange for a redrawing of the border.
I think the most important thing to understand about conflict is
that it is a _goals_ problem, not a _perception_ problem. Once you
get this, then you can see that there is no way for the system
itself to solve a conflict other than by figuring out a better
way to set the goals that are creating the conflict.
I do get this. I have never intended to say otherwise.
In the conversation/meeting conflict, the control systems inside you
that want to have a conversation and get to the meeting are setting
inconsistent goals for the system controlling where you are located
(in the conversation or at the meeting). Only a higher level system
can change the way these goals are set; this is why the sequence
control system is a solution to this conflict; in order to control the
sequence perception "conversation then meeting" the sequence control
system determines how the lower level conversation and meeting
control systems will set their goals (first one then the other).
Yes, this is perfectly clear to me.
Best Offer