[From Bruce Nevin (991112.1112 EDT)]
Marc Abrams (991112.1959)--
Me (991112.1034 EDT):
Coercion as overwhelming use of force to control while in conflict
with another control system is well understood and easily modelled.
No model has been designed, implemented, and tested of coercion by
threat. This is because no model has been designed, implemented,
and tested of communication.
Marc:
Any ideas on where to begin?
You have to figure out symbolization before you can model communication.
Martin has some ideas about communication in his "layered protocol" writings.
Threat and blandishment are formally identical. One control system is able
to determine the disposition of some variable that the other wishes to
control. Let's call them Ace and Deuce, respectively. To communicate the
threat or the promised reward Ace in some way brings to Deuce's attention
the fact that Ace is controlling this variable in a way that Deuce cannot
counter. Ace may also need to remind Deuce that Deuce wishes to control
that variable, or perhaps awaken a desire to control it that previously had
not occurred to Deuce. This is what they call "motivating" people.
Threats and blandishments are communicated as conditions upon Deuce's
control of some entirely different variable in a way that Ace desires: If
you do x in thus and such a way (or, alternatively, if you do not), then
the threat will be carried out, or the reward will be delivered.
Except for the mystery about communication, all this is pretty
straightforward -- and all of this is from the point of view of the
dominant control system, Ace. Let's look at the subordinate control system,
Deuce.
When the threat or blandishment is communicated, Deuce must believe that it
is a true condition, a property of Deuce's environment to which Ace happens
to have called Deuce's attention. Without credibility, neither a threat nor
a blandishment is effective.
Deuce controls both terms of the conditional in imagination. This is what
it's like if I do what Ace wants (possible internal conflict, the reward
delivered or the threat averted). This is what it's like if I don't (the
threat carried out or the reward foregone). Credibility gets established
when Deuce first does this controlling in imagination. There are three
possibilities: the condition that Ace offers is credible, or it is not, or
Deuce is not sure.
If it matters little to Deuce -- it's easy to help Ace out in this matter
and there's no internal conflict about it -- then uncertainty about Ace's
credibility matters little. If it does matter, Deuce probably does some
checking. Does Ace really control that variable? Would Deuce lose a
conflict with Ace? Various scenarios can develop here.
About the variable that Ace controls--the punishment or the reward--there
is also the question of how much it matters to Deuce. Deuce compares the
imagined consequences of one choice against the imagined consequences of
the other.
Suppose Deuce complies with a threat, or complies for the sake of a
promised reward. Deuce changes the reference and/or gain for the variable x
designated by Ace. This variable x is a different variable from the
inducement (the threat or reward variable). If Deuce already was
controlling that variable x with some value different from the value that
Ace wants (Deuce's new reference value), that puts Deuce in internal
conflict at the level of whatever control system(s) set that original
reference; likewise if Deuce wants the same thing now, but Deuce's
internally generated reference diverges in the future from the value that
is tied (through communication) to the threat or reward, the value that Ace
wants. If Deuce doesn't care about that variable, there is no internal
conflict. I agree with Bill that if Deuce does care, the values will
eventually diverge and there will be either internal or external conflict.
Deuce may also control a higher-level variable something like "I should be
controlling my inputs according to my own lights, thank you very much" that
would result in conflict even where the lower-level variable itself is a
"don't care". This seems to be an important bit of self-concept in some
people, less so in others. Some people are more like dogs, some more like
cats. In addition, its prevalence in populations seems to be a cultural
variable, or perhaps it is only the ability to lower the gain on it that is
a cultural variable, for the sake of learning from an expert, say.
A start might be to model parts of the above interaction without involving
communication. You would need to re-cast it with only one control system
together with certain properties of its environment. The control system
must control a variable in a not-desired state in order to avoid
anticipated undesired values of some other variable, where the effect of
one variable on the other is mediated through insentient cause-effect
chains in the environment rather than by another autonomous control system.
I don't think that has been done. Maybe it requires control in imagination
followed by control closed through the environment as I have speculated
above. The fact that the control system controls the consequence in
imagination -- imagines incurring it by action and avoiding it by inaction
(or the converse, as the case may be) -- and then pursues the most desired
alternative (or the least feared one, as the case may be) is a measure of
credibility: is that imagined consequence a true property of the
environment or not?
To model the effect of one variable on another mediated through a
conflicting autonomous control system (where one variable is the subject of
a threat or promise, and the other is the contended variable) requires
communication of the threat. Once the threat has been communicated and
believed, the above would do, I think. But I could be wrong.
Bruce Nevin
···
At 07:59 PM 11/12/1999 -0500, Marc Abrams wrote: