[From Bill Powers (960129.0800 MST)]

Hans Blom, 960129b --

Rick Marken says:

I think we (PCTers and quasi-PCTers) all agree that the environment
(or our experience thereof) does _not_ control our behavior.

and you ask:

     What is the definition of "control" that is both consistent with
     this remark and with the fact that in the PCT hierarchy (lower
     level) references have values that depend upon (perceptions of) the

OK, let's just drop the word "control." Here is a new word: retrofaction
(noun), or retrofact (verb). An active agent A is said to retrofact a
quantity or variable B if, for every disturbance tending to alter B, A
produces an action on B that tends strongly to restore B to a particular
state. For retrofaction to occur, it must be shown that this counter-
action depends on A receiving information about the state of B (more
details could be added but this will suffice for now). A formal Test for
the Retrofacted Variable could be devised; actually, it already exists.

Now, to show that the environment (A) retrofacts the behavior of an
organism (B), you must demonstrate two things. First, for any
disturbance tending to cause a change in the organism's behavior, the
environment must produce an action that tends strongly to restore the
behavior to a particular state. Second, this effect depends on the
environment's receiving information about the state of the behavior.

This is what retrofaction means. It means NOTHING ELSE. It does not mean
"influence" or "determine" or "affect" or "cause" or (now) "control." A
variable that is retrofacted by a system's actions does not "depend" on
those actions. Retrofaction means only and exactly what the definition
says it means, and it never means anything else.

So now we may ask, do organisms retrofact variables in their
environment? The answer is yes, they do. This can be demonstrated by
applying the definition to well-described observations of the
relationship between an organism and appropriate environmental
variables. We can also ask, do physical environments retrofact the
behavior of organisms? And the answer is no, they do not. If a
disturbance changes an organism's behavior, the environment does not
produce an action tending strongly to restore the behavior to a
particular state -- unless the environment happens to contain a natural
or artificial retrofactive system that has this explicit function.

By further analysis we can show that behavior actually retrofacts
perceptions of environmental variables. By applying disturbances at
various places along the link between action and input, we can determine
that only the input information is retrofacted; by exploring the nervous
system we can go further and show that it is only the signals
representing this input information that are retrofacted. So for short,
we can say that behavior is the retrofaction of perception.

We can still speak of input functions, perceptual signals, comparators,
reference signals, and output functions. We can speak of a hierarchy of
retrofactive systems. The only difference will be that now we describe
the operation of a system organized like this as retrofaction, not
control. We will bow to the inevitable: people are simply going to go on
using "control" indiscriminately to mean not only retrofact, but
influence, determine, affect, cause, and so forth -- quite different
processes. They will probably not make consistent distinctions among the
meanings of these other terms, either, so in speaking of natural
processes they will mix them all together into a fuzzy blob and think
they are talking about something -- as usual.

Of course we can't really win this battle, either. What people will do
is to hear the word "retrofaction" and immediately convert it into the
nearest term they are accustomed to using, paying no attention to our
formal definition and quite likely being unwilling to admit that there
is a phenomenon of which they have been unaware. But perhaps for a
little while we will have a vacation from people asking, "But don't
environments retrofact behavior just as much as behavior retrofacts

And perhaps some people might even notice that there's a difference
between retrofaction and the kinds of systemic behaviors they are used
to thinking about. PRT forever!



Bill P.

[Martin Taylor 960129 16:50]

Bill Powers (960129.0800 MST)

We can also ask, do physical environments retrofact the
behavior of organisms? And the answer is no, they do not.

You're going to hate me again :slight_smile:

The behaviour of the organism is to launch itself upward from the
more or less constant altitude at which it finds itself. Does the
environment retrofact this behaviour? Yes it does, quite accurately.

I think you have to show that your second criterion:

Second, this effect depends on the
environment's receiving information about the state of the behavior.

is false, and how are you going to prove a negative? Just because you
can't see an organism's-altitude-sensing-system in the environment
doesn't mean there isn't one.

Hans seems to see "control" as "bringing to a predeterminable state" or
something like that. You have a clear view of the necessary constituents
of a control loop, which include both a sensor and a power source that
is independent of the power contributed (if any) by the disturbance.
Gain is required. But how to define "retrofaction" in a way that conveys
what you want and nothing else, using only external observation, is
a bit tricky.