Controlled and unconrolled behaviour (was Re: "mental typo")

[MartinT]

[Martin Taylor 2014.01.02.23.00]

[From Fred Nickols (2014.01.02.1556 EST)]

Uh-oh. I think I'm about to commit heresy here.

No heresy unless our loose canon decides it is.

I agree that PCT explains "observed actions of animals" (and other life
forms as well). Where I might be a wee bit heretical is with regard to the
PCT canon that "behavior is the control of perception." I appreciate the
sentiment behind that statement but I would argue that it isn't technically
correct or particularly useful. I do agree that behavior serves to control
perception. No quarrel there. However, I also happen to believe (along
with millions of others) that behavior is a term that, in general, refers to
the activity of organisms (or, as Ted says, the "observed actions of
animals").
So, PCT is a theory. That theory explains behavior. PCT is not behavior.
The control of perception is not behavior, although behavior does indeed
serve to control perception.

PCT is based on an assumption that is both intuitively reasonable (that one acts to make one's world seem good and stay that way), and that is thermodynamically required (that the maintenance of structure requires avoidance of an compensation for the influences of the environment). The PCT mantra is "All behaviour is the control of perception", where "perception" means the value of some variable internal to the organism. The word "All" is crucial here, but the difficult word is "behaviour". What does it mean, operationally? Is it "behaviour" to canvass for a political party? To tense a muscle? To try to win an argument? To take a holiday in a foreign land? In casual speech, the answer is "yes" to all of these. But that's not good enough for a precise theory.

One aspect of control that is usually ignored on CSGnet is the side-effect. When you talk about "the observed activities of animals" you always mean "the effects of the outputs of the animals on the part of the environment you observe". You don't mean their actual outputs, which you can't observe, and you don't mean the intended effects of their outputs on their environment. The actual effects of someone's output on their environment may well be quite different from what they intended. When one skis, for example, one does not intend to break a leg, but that may be the result of an intention to have an exhilarating slide down a mountainside.

Is "breaking a leg" while skiing "behaviour"? Is the perfect glide while enjoying the mountain scenery? Are the various muscle tensings and relaxings that happen during either of those events?" What IS "behaviour"? Is it the intended effects on the environment, meaning the influence that counters disturbances and brings perceptions to and maintains them near their reference levels? Those effects may not happen, as when a novice tightrope walker falls into the safety net. Which is the behaviour, the walk or the fall, or both? It's not an easy question.

If we focus on the intention behind the action -- the reference value for a perception in an active control loop--, "behaviour" must consist of all the results of the output of that control loop, including the side-effects. In that case, only outputs that are independent of any control loop can be considered "uncontrolled behaviour". The glass unintentionally knocked over when reaching for something else is not "uncontrolled behaviour" but "unintended consequences of behaviour used in control of some other perception". It's a side-effect. All observable behaviour has side-effects. The fact that the behaviour can be observed is proof of that.

Can there be "uncontrolled behaviour" within the HPCT construct? Yes, there can. It can happen at least when the brain or a motor nerve is externally stimulated. At some level, the external stimulation generates an output in some system independently of the levels of the reference and the perception in that system. At lower levels, the behaviour (outputs) are part of functioning control loops, but at the level of the stimulation the output is independent of control. It is a pure input->effect on the environment lineal causal system.

Side-effects of control may be seen as lineal causal effects of output

···

on the environment. In that sense they are "uncontrolled", but then the thesis of PCT is not that we control behaviour, is it? The side-effect of action is neither more not less controlled than any other aspect of action. Perception is what is controlled, not output. Output may be said to be "commanded", but it isn't "controlled", at least not within any flavour of PCT. Any side-effects are just among the changes in the environment that occur as consequences of the outputs that influence the perception.

No observer can tell which effects on the environment are side-effects and which are necessary to influence the controlled perceptions. When the glass is knocked over, you can't use the TCV to test whether the spill was intended. It happened, the social and physical circumstances have changed irrevocably, and the Test cannot be performed. The TCV can be used sometimes, but in real life it often is impossible. The effect on the environment occurred, the situation and opportunities to influence perceptions have changed, and no observer can distinguish side-effect from intentional effect other than by emulating the TCV imperfectly in imagination, using analogous remembered situations.

Fred says: "The control of perception is not behavior, although behavior does indeed serve to control perception." So we close the circle and ask again " what could be meant by 'controlled' and 'uncontrolled' behaviour"? My answer now is that all behaviour is uncontrolled. The distinction intended by using the words is between intended and side effects on the environment. Would it not be better to use those terms instead of implying that some behaviour is controlled?

Martin

[rsmarken]

[From Rick Marken (2014.01.04.1110)]

···

Martin Taylor (2014.01.02.23.00)

MT: The PCT mantra is “All behaviour is the control of perception”

RM: Not “all behavior” is the control of perception; only behavior that is intentional (purposeful), ie. behavior that is control.

MT: Is “breaking a leg” while skiing “behaviour”?

RM: Yes, but it is not control unless it was broken intentionally.

MT: If we focus on the intention behind the action – the reference value for a perception in an active control loop–, “behaviour” must consist of all the results of the output of that control loop, including the side-effects.

RM: If we focus on intended actions (or any other intended results, including the consequences of those actgions) then we are focusing only on behavior that is control. Side-effects of control are unintended actions/results.

MT: Side-effects of control may be seen as lineal causal effects of output on the environment. In that sense they are “uncontrolled”,

RM: Correct; they are uncontrolled variables.

MT: but then the thesis of PCT is not that we control behaviour, is it?

RM: Right. The thesis (observation, really) of PCT is that behavior is control, not that we control behavior.

MT: No observer can tell which effects on the environment are side-effects and which are necessary to influence the controlled perceptions.

RM: Yes, but an observer can tell which effects are side-effects (uncontrolled variables) and which are controlled (controlled variables) by doing the TCV.

MT: When the glass is knocked over, you can’t use the TCV to test whether the spill was intended.

RM: Yes, that’s true. But the determination of whether and what is intended (ie. whether a particular result is controlled) is typically done under laboratory conditions. You can’t determine controlled variables “in the field” any better than you can determine the laws of motion by watching a baseball game.

MT: So we close the circle and ask again " what could be meant by ‘controlled’ and ‘uncontrolled’ behaviour"?

RM: The question I have been addressing is quite different. I have been addressing the question of distinguishing behavior that is control from behavior that is not. Behavior that is control is seen when a system is “acting to bring something to a specified condition, and then maintaining
it close to that condition even if unpredictable external forces and changes in environmental properties tend to alter it”. Behavior that is not control is everything else that we see the system doing.

MT: My answer now is that all behaviour is uncontrolled.

RM: But you have not said what you mean by behavior (or behaviour if you prefer). I define behavior as “control” (even though some things we call behavior are side effects of control and are, thus, not control) because that’s the aspect of what is called “behavior” that control theory (PCT) explains. And it is manifestly possible to control behavior that is control (see my demonstration thereof at http://www.mindreadings.com/ControlDemo/Coercion.html)

Best

Rick

–
Richard S. Marken PhD
www.mindreadings.com
The only thing that will redeem mankind is cooperation.

                                               -- Bertrand Russell

[MartinT]

[Martin Taylor 2014.01.04.14.35]

Yes, but to say that is to imply a definition of "behaviour", which

is problematic. That’s what I was addressing much of my message.
Later, you make this definition circular, or so it seems to me.
“Breaking a leg”, or any other consequence of action, is not
control. Control is a process that takes place in a loop. No
fragment of that loop is in itself “control”.
You are right, but you miss my point. I’ll try to put it in a
different way.
All actions by a biological entity have effects on the environment
of the entity. Not all of these effects may be observable by an
outside observer, but of those that are, all might be called the
“behaviour” of the entity. The acting entity may know which effects
are intended and which are not, but no observer can know it without
making assumptions that may or may not be correct.
Some of those effects are essential if the environment is to be
changed in ways that affect the controlled perception(s). Some are
not. Even those that are might be called “side-effects”; for
example, a “side-effect” of an archaeological dig is to destroy the
structure of the overlying material, but the archaeologist cannot
see the main target of the investigation without doing so. Almost
all action changes the environment in ways that are essential to the
purpose, and that change the way the environment can be used for
other purposes by the same or a different actor. They are part of
the feedback loop, and are intended effects insofar as they are
parts of the feedback loops in the control of lower-level
perceptions. They are part of “the behaviour” of controlling the
perception whose control is the main purpose.
But one of (to me) the great benefits of understanding the
generality of PCT is that it can be used in the field where the
results of conventional lab studies are often misleading when they
are extrapolated beyond the lab situation. Although one cannot use
the TCV to determine what perceptions were controlled in a one-shot
observed series of events, one can make assumptions that are often
correct, as determined by a loose form of the TCV. I think of an
example from my graduate student days.
At the time, I lived in a house with three other graduate students
(all in different disciplines). One, I’ll call him “Bob” was what we
called “accident prone” (a popular term at the time). Now
“accident-prone” could reflect a perfectly expectable statistical
effect. Some people just get into more accidents than others. But
observing Bob suggested a different answer. We washed up by hand in
the sink. Bob would put all the cutlery in the sink with the points
of kitchen knives pointing upward, and later would plunge his hand
into the water without looking and get cut. He would place something
on the stairs, and later trip over it. Month after month, one could
see him setting up “accidents”, which sometimes one could forestall
by rearranging his environment, but he still managed to find ways to
have accidents we hadn’t noticed him arrange. I’m quite sure Bob did not set up his accidents with conscious
intention, but in the PCT sense he did set them up with purpose, and
I can hazard a guess where that reference came from. Bob was a
Mormon, and Mormons are supposed to do time as missionaries when
they are called to do so. He didn’t want to leave his studies, and
we had quite a game making sure he was not visible when people came
to the door who looked as though they might be Mormon missionaries
coming to give the call (two young men in dark suits, as I
remember). I am now guessing that Bob perceived himself as requiring
punishment for evading his heavenly duty, a perception achieved by
way of his accidents. The trouble with such guesses in real life is that they can be
wildly wrong, but if they are correct, it may be possible to
determine whether some behaviours are controlled. If Bob, for
example, had “accidentally” knocked over the proverbial glass, the
accident might have been part of a control loop that served to
provide him with the disapproval of his host, a mild punishment. But
you seldom can be sure, even if the actor specifically tells you
whether something was intentional and why it was done.
I don’t think that’s a different question. You are trying to
distinguish the same thing I am, except that you see control as a
synonym for one kind of behaviour, whereas I think of controlled
behaviour as forming only the observable part of a control loop. But
since ALL side-effects are a direct consequence of the activity of
control loops, there is no such thing as uncontrolled behaviour (or,
to use your words, behaviour that is not control). However, there is
such a thing as unintended environmental effects of behaviour.
I think that “everything else” is very close to an empty set, and is
driven only by neural noise and such random fluctuations. All the
biologically caused effects in the world are effects caused by
"“acting to bring something to a specified condition, and then
maintaining it close to that condition even if unpredictable
external forces and changes in environmental properties tend to
alter it”.
No, I haven’t. And I won’t. The reason I haven’t is that I don’t
think that it’s possible to generate a precise definition that is
better than the everyday one: “behaviour is what you can see someone
doing”. Since you can’t see the intention or assess what is being
controlled, there’s no point in trying to distinguish kinds of
behaviour as control or as not control. The second mantra of PCT,
after “All behaviour is the control of perception”, is “You can’t
see what someone is doing by looking at what they are doing”. I
believe in both. Behaviour is the external manifestation of control,
but you can’t tell which observable effects are part of that control
and which are not (other than in the laboratory, where you assume
the subject is taking on some reference value you have provided).
Yes it is, if you can correctly find some perception someone is
controlling, find a way to disturb it, and arrange that there is a
high probability that the person being controlled will use a
particular set of actions to counter the disturbance. We do it every
day, but we usually don’t call it coercion. We call it civility,
politeness, mutual goodwill.
Anyway, I simply wanted to point out that side-effects are
unfortunately given short shrift on CSGnet, despite being rather
important in their effects on the environment and upon other actors
who can perceive that environment.
Martin

···

For once, I think Rick and I are pretty much on the same page,
though perhaps not on the same part of the same page.

[From Rick Marken (2014.01.04.1110)]

Martin Taylor (2014.01.02.23.00)

          MT: The PCT mantra is

“All behaviour is the control of perception”

          RM: Not "all behavior" is the control of perception;

only behavior that is intentional (purposeful), ie.
behavior that is control.

MT: Is “breaking a leg” while skiing “behaviour”?

          RM: Yes, but it is not control unless it was broken

intentionally.

          MT: If we focus on the intention behind the action -- the

reference value for a perception in an active control
loop–, “behaviour” must consist of all the results of the
output of that control loop, including the side-effects.

          RM: If we focus on intended actions (or any other

intended results, including the consequences of those
actgions) then we are focusing only on behavior that is
control. Side-effects of control are unintended
actions/results.

          MT: Side-effects of

control may be seen as lineal causal effects of output on
the environment. In that sense they are “uncontrolled”,

RM: Correct; they are uncontrolled variables.

          MT: but then the

thesis of PCT is not that we control behaviour, is it?

          RM: Right. The thesis (observation, really) of PCT is

that behavior is control, not that we control behavior.

          MT: No observer can tell which effects on the environment

are side-effects and which are necessary to influence the
controlled perceptions.

          RM: Yes, but an observer can tell which effects are

side-effects (uncontrolled variables) and which are
controlled (controlled variables) by doing the TCV.

          MT: When the glass is

knocked over, you can’t use the TCV to test whether the
spill was intended.

          RM: Yes, that's true. But the determination of whether

and what is intended (ie. whether a particular result is
controlled) is typically done under laboratory conditions.
You can’t determine controlled variables “in the field”
any better than you can determine the laws of motion by
watching a baseball game.

          MT: So we close the

circle and ask again " what could be meant by ‘controlled’
and ‘uncontrolled’ behaviour"?

          RM: The question I have been addressing is quite

different. I have been addressing the question of
distinguishing behavior that is control from behavior
that is not.

          Behavior that is control is seen when a system is

“acting to bring something to a specified condition, and
then maintaining it close to that condition even if
unpredictable external forces and changes in environmental
properties tend to alter it”. Behavior that is not control
is everything else that we see the system doing.

MT: My answer now is that all behaviour is uncontrolled.

          RM: But you have not said what you mean by behavior (or

behaviour if you prefer).

          I define behavior as "control" (even though some things

we call behavior are side effects of control and are,
thus, not control) because that’s the aspect of what is
called “behavior” that control theory (PCT) explains. And
it is manifestly possible to control behavior that is
control (see my demonstration thereof at http://www.mindreadings.com/ControlDemo/Coercion.html)

[rsmarken]

[From Rick Marken (2014.01.05.1200)]

Martin Taylor (2014.01.04.14.35) --

MT: Is "breaking a leg" while skiing "behaviour"?

RM: Yes, but it is not control unless it was broken intentionally.

MT: "Breaking a leg", or any other consequence of action, is not control.

RM: Consequences of action that are protected from disturbance are
controlled (intentionally produced) consequences of action. If one's
intention is to break one's leg then "breaking a leg" is a process of
control and the broken leg is a controlled consequence of the
combination of actions and independent environmental effects
(disturbances) that produce that result.

MT: The acting entity may know which effects are intended and which
are not, but no observer can know it without making assumptions that
may or may not be correct.

RM: All that is needed to discriminate intended from unintended
consequences of action is to look at the effect of disturbances to
those consequences. If the effect of the disturbances is what is
expected based on physical law then the consequence is not intended
(not controlled). If the effect is _not_ what is expected then the
consequence is likely to have been produced intentionally
(controlled). This is demonstrated by my Mind Reading demo
(Mindreading) as

MT: But one of (to me) the great benefits of understanding the generality of
PCT is that it can be used in the field where the results of conventional lab
studies are often misleading when they are extrapolated beyond the lab
situation.

RM: Yes, PCT can certainly be used in the field but the results are
not as precise as those that you get in the lab. If one goal of the
field research is to determine what variable(s) an organism is
controlling you have to either look for naturally occurring
disturbances or produce some disturbances yourself. Such an
intervention was apparently done in a field study of beaver dam
building behavior. The researcher wanted to know what variable(s) the
beaver controls by building the dam. One hypothesis was the sound of
rushing water, trying to keep the sounds level at zero. So the
researcher played a recording of rushing water after a dam was
completed and the beavers apparently worked to "dam up" the speaker,
covering it with mud to stop the water flow noise.

MT: The trouble with such guesses in real life is that they can be wildly
wrong, but if they are correct, it may be possible to determine whether
some behaviours are controlled.

RM: Variables are controlled. Behavior is control.

RM: Behavior that is control is seen when a system is "acting to bring
something to a specified condition, and then maintaining it close to
that condition even if unpredictable external forces and changes in
environmental properties tend to alter it". Behavior that is not control is
everything else that we see the system doing.

MT: I think that "everything else" is very close to an empty set, and is
driven only by neural noise and such random fluctuations.

RM: See the "Squaring the Circle" demo on p. 145-150 in LCS III. The
circular path of the mouse is one unintended side effect of
controlling for making the cursor move in a square path on the screen.
There are actually many such unintended side effects that are not
recorded in this demo -- such as rotational movements of the mouse,
sounds made when moving the mouse, etc -- but the circular path traced
out by the mouse while making a square pattern with the cursor is
certainly the most dramatic -- and the most likely to be seen as an
intended behavior itself.

MT: I don't think that it's possible to generate a precise definition
[of behavior] that is better than the everyday one: "behaviour is
what you can see someone doing". Since you can't see the intention
or assess what is being controlled, there's no point in trying to
distinguish kinds of behaviour as control or as not control.

RM: You can get access to the controlled quantity, which is the basis
of the controlled perception, because it is an aspect of the system's
environment. For example, the distance between cursor and target, c -
t, is one possible controlled quantity (the basis of the controlled
perceptual variable) in a tracking task. By applying disturbances to
this variable and seeing that they have far less than the expected
effect you can infer that the person's intention in the task is to
control a perception of this quantity. Of course, we found that a
better definition of the controlled quantity is the angular separation
between cursor and target; in other words arctan (c-t/s), where s is
the separation between c and t that is orthogonal to c-t. So arctan
(c-t/s) is a better guess about the perception controlled in a
tracking task than c-t.

MT: The second mantra of PCT, after "All behaviour is the control of
perception", is "You can't see what someone is doing by looking at
what they are doing". I believe in both.

RM: Great. Then you know that you _can_ tell what a person is "doing"
intentionally (that is, what they are controlling) by nit just looking
but by actively manipulating variables that should have effects on
other variables (hypothetical controlled variables) if they are not
under control and looking for lack of effect. So another "mantra" of
PCT is that you can see what someone is doing by testing to determine
what variable(s) they are controlling"

MT: Anyway, I simply wanted to point out that side-effects are unfortunately
given short shrift on CSGnet, despite being rather important in their
effects on the environment and upon other actors who can perceive that
environment.

RM: Bill gives them pretty "long shrift" in the "Squaring the Circle"
section of LCS III. A brilliant demo!

Best

Rick

···

--
Richard S. Marken PhD
www.mindreadings.com

The only thing that will redeem mankind is cooperation.
                                                   -- Bertrand Russell

[rsmarken]

[From Rick Marken (2014.01.05.1310)]

I seem to have accidentally deleted something from my earlier post.
Here's the completed paragraph:

Rick Marken (2014.01.05.1200)--

RM: All that is needed to discriminate intended from unintended
consequences of action is to look at the effect of disturbances to
those consequences. If the effect of the disturbances is what is
expected based on physical law then the consequence is not intended
(not controlled). If the effect is _not_ what is expected then the
consequence is likely to have been produced intentionally
(controlled). This is demonstrated by my Mind Reading demo
(Mindreading) as
we as my less well know (but probably easier to understand )
Detection of intention demo (Detection of Intention)

Best

Rick

···

--
Richard S. Marken PhD
www.mindreadings.com

The only thing that will redeem mankind is cooperation.
                                                   -- Bertrand Russell

[MartinT]

[Martin Taylor 2014.01.06.00.02]

[From Rick Marken (2014.01.05.1200)]

Martin Taylor (2014.01.04.14.35) --

MT: Is "breaking a leg" while skiing "behaviour"?

RM: Yes, but it is not control unless it was broken intentionally.

MT: "Breaking a leg", or any other consequence of action, is not control.

RM: Consequences of action that are protected from disturbance are
controlled (intentionally produced) consequences of action. If one's
intention is to break one's leg then "breaking a leg" is a process of
control and the broken leg is a controlled consequence of the
combination of actions and independent environmental effects
(disturbances) that produce that result.

The actor may (or may not) know whether it was intentional. Nobody else does.

As for the rest of your comments, I think they are not unlike answers in Question Period in our House of Commons (Parliament -- it's a daily session in which Government Ministers are supposed to answer questions from the Opposition): answers consist largely of restating tired talking points that have little relevance to the topic of the question, and are designed to make Her Majesty's Loyal Opposition look bad.

MT: Anyway, I simply wanted to point out that side-effects are unfortunately
given short shrift on CSGnet, despite being rather important in their
effects on the environment and upon other actors who can perceive that
environment.

RM: Bill gives them pretty "long shrift" in the "Squaring the Circle"
section of LCS III. A brilliant demo!

Maybe so, but CSGnet doesn't consist of quotes from LCSIII.

Incidentally, I remember Bill once saying a long time ago that up to that point I was the only one who had really seen the value of the Squaring the Circle demo, a demo which I agree now, and said at the time, is brilliant.

Martin