A PCT analysis of violence

[From Bill Powers (940526.0700 MDT)]

Oded Maler (940526) --

Thanks for relaying the welcome-to-me information about Thigh Cream.
Looks as though there are plenty of people on the internet who
consider unwanted ads a disturbance of an important controlled
variable. A little resistance by a lot of people adds up to a brick
wall -- as we in PCT know from our experiences with trying to be a
disturbance (or at least turning out to be one, whether we wanted to
or not).

:So everybody can relax and preach non-violent control as usual :slight_smile:

I think there's some misunderstanding about "preaching nonviolent
control." There aren't two modes of control, violent and nonviolent,
which are alternatives. There is only control, the process of acting
to keep the perceived world as close as possible to a match with the
states one desires to perceive.

When a controlled variable departs from its specified state, the
resulting error produces actions that tend to change the controlled
variable back toward the specified state. The greater the departure,
the greater the amount of action. That is simply how control works.

There are two main determinants of the amount of action that will
appear for one unit of departure. One is the loop gain, which we can
define as the apparent "importance" of the controlled variable. The
other is the amount of effect that one unit of action has on the
controlled variable: that is, the tightness of the coupling between
the action and the controlled variable. If one unit of action has
only a small effect on the controlled variable, there will be a
large amount of action; if a larger effect, a smaller amount of
action. Given sufficient loop gain, the action will in any case be
large enough to oppose whatever disturbance is tending to change the
controlled variable.

If a second control system acts in a way that disturbs the
controlled variable of the first system, some degree of conflict
will develop. The amount of conflict depends on the coupling between
the controlled variables of the respective control systems, with the
maximum potential for conflict arising when the controlled variables
are identical. A conflict results in the actions of two control
systems changing to oppose each other, each system reacting to the
actions of the other system as a disturbance.

When the loop gains are low, the result may be some compromise state
of the controlled variable, with significant error existing in both
systems but neither system exerting a large amount of action against
the other. We could say that controlling this variable is not
important to either system, measuring importance in terms of the
apparent loop gain.

When the loop gain is low in one system and high in the other, the
system with the larger loop gain will experience the smaller error.
This is the case where control of the common variable is highly
important to one system and less important to the other.

When the loop gain is high in both systems (the controlled variable
is highly important to both systems), the level of action will be
correspondingly high, and neither system will be able to correct its
error. A high level of action is what we call "violence."

If the controlled variable is not very important to either system,
there is no need to resolve the conflict: there is error, but is it
not enough to produce any significant effort to correct it.

When one system considers the controlled variable to be very
important but the other does not, the system treating the variable
as least important will experience the largest error. That system
will probably be the one that changes goals, seeking some other
means of achieving higher-level goals and thus removing the
conflict. The amount of action called for will probably not rise to
the level that we would call violence, and as soon as the one system
changes to a different means of control, the level of action of both
systems will drop to whatever is normal without conflict, with only
natural disturbances acting.

Only when both systems consider the controlled variable important,
and neither changes its goal, will the degree of action rise to the
point where we would call it violent. As each system raises its
amount of action to force the controlled variable back to its
preferred state, the result is simply to increase the disturbance of
the other system, which raises its amount of action by the same
amount. The natural result is a large amount of action being
produced by each system against the action of the other system,
which is exactly what we mean by violence.

In the case of the Branch Davidians vs the ATF troops, the
controlled variable at the focus of the conflict was possession of
an arsensal by the Branch Davidians. The ATF wished for that arsenal
to be surrendered; the Branch Davidians wished to keep it. Both
sides assigned maximum importance to maintaining control of this
variable. Neither side was willing to change the pertinent goal. The
result was a standoff for a time, but the integral error terms on
the ATF side soon required raising the level of action, and that
raised the level of opposing action until first there was a fire-
fight, and eventually the building was stormed and the Branch
Dividians were burned up. This result was inevitable as long as
neither side was willing to change the goal regarding possession of
the arsenal.

As Kent McClelland showed in his simulations of conflict at the last
CSG meeting, the existence of a limit on the possible output of one
system in a conflicted situation can determine the final outcome,
particularly in systems with integrating output functions. When the
level of conflict has risen sufficiently, one control system is no
longer able to increase its output more to oppose further increases
in the output of the other system. From there on, the other system
wins the conflict. In the case of the Waco catastrophe, the Branch
Davidians ran out of output first. The ATF, with its superior
resources, won by virtue of superior force (not brains or
rightness).

The origins and resolution of conflict can be understood only in the
context of a hierarchical model. The conflicting goals do not appear
out of thin air; they are set as they are as a means of controlling
higher-level variables. The Branch Davidians wanted to keep their
arsenal because they perceived themselves to be under a threat of
annihilation: Armageddon, they believed, was near. Their only means
of survival lay in their stockpile of arms and ammunition; to give
it up would be suicidal. The presence of the ATF troops only served
to support their beliefs. Thus it was highly important to them to
keep control of the arsenal.

On the other side, removing the arsenal was simply a means of
maintaining law and order. The goal of making the Branch Davidians
give up their arms was chosen as a means of maintaining a higher-
level controlled variable. As long as the arms remained inside the
compound and out of the ATF's control, the threat to law and order
persisted. Thus the goal of getting control of the arsenal was very
important to the ATF, too.

The conflict was therefore _expressed_ at the level of control of
the arsenal. But it was _caused_ by the higher-level goals on both
sides that required setting the immediate goal of possessing the
arsenal. The conflict could not be RESOLVED at the level of trying
to force one side to give up the arms. It could only be won or lost
at that level. To resolve it, the two sides would have had to
discuss the higher-level perceptions and goals, and attempt to find
a compromise that was not perceived as threatening survival on the
one hand, or law and order on the other hand. The question of
possession of the arsenal might, for example, have been postponed.
Hostages might have been exchanged. Guarantees might have been at
least given, if not totally believed.

But the required flexibility at the higher levels was not there on
either side. The violent end was, as a result, unavoidable.

···

----------------------------------------------------------------
There is more that can be said without implying either a liking for
violence or a dislike of it. For example, conflict between equal
parties reduces the ability of both parties to defend the conflicted
variable against external disturbances, especially when the output
functions of both parties saturate at the same level of output. In
nonlinear systems, conflict will tend to drive both systems into
nonlinear regions where dynamic stability may be lost and
spontaneous oscillations may arise. For both systems, the existence
of conflict creates a severe drain on resources, making other modes
of control less effective. And of course, for both systems the fact
that neither can freely control the variable in contention means
that that variable is not usable by higher systems to control
higher-level perceptions.

Also, winning a conflict rather than resolving it leaves the higher-
level organization of the losing side exactly as it was, so the
loser will keep looking for other ways of achieving the higher-level
goals. If those goals are incompatible with the higher-level goals
of the winner, more conflict is very likely to arise. So winning
conflicts is only a temporary solution, if a solution is wanted.

All this has come out of a simple and straightfoward analysis of
conflict in terms of a simple control model. No preaching against
violence is involved: we are merely working out what violence is,
what leads to it, and what is required if a resolution is wanted.
What one thinks about the desirability of violence is beside the
point.
---------------------------------------------------------------
In the case of Mr. Thigh Cream, my loop gain was high but the
coupling between my actions and the controlled variable was very
weak, even nonexistent. All I could do was to keep thinking of more
outputs that might have an effect. As far as I could tell, I was not
affecting Mr. Thigh Cream at all. Today, it turned out that someone
else defending a similar controlled variable found an action that
did couple to the controlled variable, and in fact cut off the
disturbance altogether. Since I no longer anticipate the same
departure of that controlled variable from its reference level,
there is no need for more action on my part -- until another
disturbance of the same kind shows up.
---------------------------------------------------------------
A final remark. There are two kinds of violence: natural and
pathological. Natural violence arises from conflicts between
important goals. It is always two-sided and is suspectible to
resolution more often than not. Pathological violence, however, is
one-sided: it is an unprovoked attack done for the sake of creating
violent effects, the goal being only to achieve those effects. There
is no way to negotiate a resolution of such violence; all that can
be done is to contain it by superior force. If, of course, you
object to it.
---------------------------------------------------------------
Best,

Bill P.

[From Oded Maler (940526) II]

* [Bill Powers (940526.0700 MDT)]

[...] (To which I cannot respond immediately)

* A final remark. There are two kinds of violence: natural and
* pathological. Natural violence arises from conflicts between
* important goals. It is always two-sided and is suspectible to
* resolution more often than not. Pathological violence, however, is
* one-sided: it is an unprovoked attack done for the sake of creating
* violent effects, the goal being only to achieve those effects. There
* is no way to negotiate a resolution of such violence; all that can
* be done is to contain it by superior force. If, of course, you
* object to it.

Important to whom? What objective judgement can you assign to someone
else's reference signals? It is important for the pathological to
perceive the suffers of others.

--Oded

···

--

Oded Maler, VERIMAG, Miniparc ZIRST, 38330 Montbonnot, France
Phone: 76909635 Fax: 76413620 e-mail: Oded.Maler@imag.fr

[From: Oded Maler (940526)]

Re: Bill Powers (940526.0700 MDT)

[..] (Nothing to disagree with).

I just want to note that when the control systems are more complex,
and when they are embedded in a social/organizational context, the
meaning of energy and loop-gain are not as simple. You can say to
your wife something very nasty, or insult someone in public, that
will have much stronger effect then, say, mild pushing which is
more costly energetically. The same goes for social organization:
you can order the killing/execution of a person by a phone call or
by signing a document if you are in the correct position.

--Oded

···

--

Oded Maler, VERIMAG, Miniparc ZIRST, 38330 Montbonnot, France
Phone: 76909635 Fax: 76413620 e-mail: Oded.Maler@imag.fr