The "Law of Effect" Restated

CSGnet Archive CSG2012
1

[From Richard Pfau (2012.11.29 14:07 Nepal)]

A principle of psychology called the Law of Effect "holds that responses to stimuli that produce a satisfying or pleasant state of affairs in a particular situation are more likely to occur again in a similar situation. Conversely, responses that produce a discomforting, annoying or unpleasant effect are less likely to occur again in the situation." (from Wikipedia: http:/en.wikipedia.org/wiki/Law_of_effect, 2 April 2009). Other definitions and descriptions of The Law are similar, for example, "Behavior having good consequences tends to be repeated, whereas behavior that leads to bad consequences is not repeated."

from a PCT perspective, might the Law of Effect be restated as follows:

"Behaviors that reduce or eliminate error signals are more likely to occur again in similar situations. Conversely, behaviors that produce, increase, or fail to eliminate error signals are less likely to occur again in similar situations."

I ask if this restatement is OK, since I may include it in something that I am writing.

With Regards,
Richard Pfau

2

[From Richard Kennaway (2012.11.29 15:11 GMT)]

[From Richard Pfau (2012.11.29 14:07 Nepal)]

A principle of psychology called the Law of Effect "holds that responses to stimuli that produce a satisfying or pleasant state of affairs in a particular situation are more likely to occur again in a similar situation. Conversely, responses that produce a discomforting, annoying or unpleasant effect are less likely to occur again in the situation." (from Wikipedia: http:/en.wikipedia.org/wiki/Law_of_effect, 2 April 2009). Other definitions and descriptions of The Law are similar, for example, "Behavior having good consequences tends to be repeated, whereas behavior that leads to bad consequences is not repeated."

From a PCT perspective, might the Law of Effect be restated as follows:

"Behaviors that reduce or eliminate error signals are more likely to occur again in similar situations. Conversely, behaviors that produce, increase, or fail to eliminate error signals are less likely to occur again in similar situations."

I ask if this restatement is OK, since I may include it in something that I am writing.

It looks like behaviourism rearing its undead head again. Behaviourism into PCT won't go.

The problem is in that word "situation". What is the "situation", and how does that compare with what actually determines the behaviour?

By the situation, people generally mean the circumstances surrounding the animal.

But what determines behaviour is the error, the difference between perception and reference. The reference, and the error computed from it, are part of the animal, not part of the situation. The perception is more immediately connected to the "situation", but what the experimenter thinks the situation "is", is not necessarily the situation as the animal perceives it.

Even if the reference were kept constant, there is another issue. A control system will in general control with an accuracy similar to the accuracy with which it measures the error. (It clearly can't control any better, and if it did worse, that would mean its sensor was over-engineered.) The error signal usually looks like noise, and you probably can't get at it to measure anyway. And yet that is what is actually producing the behaviour, not "the situation".

···

--
Richard Kennaway, jrk@cmp.uea.ac.uk, Richard Kennaway
Tel. 01603 593212
School of Computing Sciences,
University of East Anglia, Norwich NR4 7TJ, U.K.

3

[From Richard Kennaway (2012.11.29 15:271 GMT)]

And an addendum: the problem with "situation" applies also to "behaviour". For the example of a cat getting out of a box, what is the "behaviour" that gets the cat out? In the experimenter's eyes, it's pushing a certain lever. But on different occasions, the cat may use a different leg to push it, use different muscles in different ways, and so on. What is "the" behaviour that getting out of the box is causing to be repeated next time? How did the "good consequence" of getting out of the box select "pushing the lever" from the space of everything the cat could be described as doing, in order for that thing to be a thing that works next time?

···

--
Richard Kennaway, jrk@cmp.uea.ac.uk, http://www.cmp.uea.ac.uk/~jrk/
School of Computing Sciences,
University of East Anglia, Norwich NR4 7TJ, U.K.

4

from Fred Nickols (2012.11.29.0842)

And that is why performance specialists are careful to distinguish between actions or behaviors (e.g., pushing lever or whatever) and outcomes or results (e.g., lever pushed or gate opened).

···

Sent from my iPad

On Nov 29, 2012, at 8:34, Richard Kennaway <jrk@CMP.UEA.AC.UK> wrote:

[From Richard Kennaway (2012.11.29 15:271 GMT)]

And an addendum: the problem with "situation" applies also to "behaviour". For the example of a cat getting out of a box, what is the "behaviour" that gets the cat out? In the experimenter's eyes, it's pushing a certain lever. But on different occasions, the cat may use a different leg to push it, use different muscles in different ways, and so on. What is "the" behaviour that getting out of the box is causing to be repeated next time? How did the "good consequence" of getting out of the box select "pushing the lever" from the space of everything the cat could be described as doing, in order for that thing to be a thing that works next time?

--
Richard Kennaway, jrk@cmp.uea.ac.uk, Richard Kennaway
School of Computing Sciences,
University of East Anglia, Norwich NR4 7TJ, U.K.

5

[From Fred Nickols (2012.11.29.0850 AZ)]

Wow! That was fast. This just came in...

"Hey! It's me. The cat. It took me a while but I finally figured out that
if I can get that long piece of wood y'all call a "lever" moved from its up
position to its down position that I can get out of that darned box in which
some jerk of a human being keeps putting me. It doesn't really make any
difference which particular stance I take or where I happen to be sitting;
if I can exert enough downward pressure the lever goes down and I get out.
On the other hand, if I don't want out, I can just sit there. However, that
darned human being keeps depriving me of food so I'm generally pretty hungry
and I know there's food outside the box. So, down goes the lever and out I
go. It's no big deal, really, but don't you human beings have better things
to do?"

The Cat

From: Control Systems Group Network (CSGnet)
[mailto:CSGNET@LISTSERV.ILLINOIS.EDU] On Behalf Of Fred Nickols
Sent: Thursday, November 29, 2012 8:46 AM
To: CSGNET@LISTSERV.ILLINOIS.EDU
Subject: Re: The "Law of Effect" Restated

From Fred Nickols (2012.11.29.0842)

And that is why performance specialists are careful to distinguish between
actions or behaviors (e.g., pushing lever or whatever) and outcomes or
results (e.g., lever pushed or gate opened).

Sent from my iPad

> [From Richard Kennaway (2012.11.29 15:271 GMT)]
>
> And an addendum: the problem with "situation" applies also to
"behaviour". For the example of a cat getting out of a box, what is the
"behaviour" that gets the cat out? In the experimenter's eyes, it's

pushing a

certain lever. But on different occasions, the cat may use a different

leg to

push it, use different muscles in different ways, and so on. What is

"the"

behaviour that getting out of the box is causing to be repeated next time?
How did the "good consequence" of getting out of the box select "pushing
the lever" from the space of everything the cat could be described as

doing,

···

-----Original Message-----
On Nov 29, 2012, at 8:34, Richard Kennaway <jrk@CMP.UEA.AC.UK> wrote:
in order for that thing to be a thing that works next time?
>
> --
> Richard Kennaway, jrk@cmp.uea.ac.uk, Richard Kennaway
> School of Computing Sciences, University of East Anglia, Norwich NR4
> 7TJ, U.K.

6

[From Rickk Marken (2012.11.29.0805)

Fred Nickols (2012.11.29.0850 AZ)--

FNC: "Hey! It's me. The cat. It took me a while but I finally figured out that
if I can get that long piece of wood y'all call a "lever" moved from its up
position to its down position that I can get out of that darned box in which
some jerk of a human being keeps putting me. It doesn't really make any
difference which particular stance I take or where I happen to be sitting;
if I can exert enough downward pressure the lever goes down and I get out.
On the other hand, if I don't want out, I can just sit there. However, that
darned human being keeps depriving me of food so I'm generally pretty hungry
and I know there's food outside the box. So, down goes the lever and out I
go. It's no big deal, really, but don't you human beings have better things
to do?"

Thanks Fred and FNC (Fred Nickols Cat, of course). That was gorgeous!!

Best

Rick

···

--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

7

bob hintz 11.29.12

“Behaviors that reduce or eliminate error signals are more likely to occur again in similar situations. Conversely, behaviors that produce, increase, or fail to eliminate error signals are less likely to occur again in similar situations.”

Fred Nickols (2012.11.29.0850 AZ)–

FNC: "Hey! It’s me. The cat. It took me a while but I finally figured out that
if I can get that long piece of wood y’all call a “lever” moved from its up

position to its down position that I can get out of that darned box in which

some jerk of a human being keeps putting me. It doesn’t really make any

difference which particular stance I take or where I happen to be sitting;

if I can exert enough downward pressure the lever goes down and I get out.

On the other hand, if I don’t want out, I can just sit there. However, that

darned human being keeps depriving me of food so I’m generally pretty hungry

and I know there’s food outside the box. So, down goes the lever and out I

go. It’s no big deal, really, but don’t you human beings have better things

to do?"

BHC: I thought I had it made, but the damn lever didn’t work in this new box Bob put me in. I got so frustrated that I howled and howled - that works when I get stuck in the basement so someone comes and opens the door after awhile. But that didn’t work either. I tried pushing on the lid and scratched at each of the sides of the box. I finally noticed a ball on the wall opposite from the lever. I didn’t feel much like playing, but I swatted it any way. I was really shocked when the wall fell down and I could walk out of that damn box and get to food I had been smelling since he put me in the box. The next time he puts me in a box I will check out anything that sticks out from a wall.

Gregory Bateson once wrote about zero learning, learning 1, and learning 2. When the cat tried the level, he was engaging in zero learning. He was just using past experience to correct an error that he was experiencing. (Might the lever have been considered an effordance?) When the lever didn’t work, he needed to learn something new and had to try a variety of different behaviors if he wanted to get out of the box. When he discovered the new devise through trial and error, he succeeded in an instance of learning 1 and had a new option for escaping from boxes that humans seem determined to put him in. In the next box, he might try the ball first if there is one. However, when he formulated a “plan” for learning how to escape from the next box he has a learning how to learn strategy which Bateson labeled learning 2. If the learning 2 strategy fails, he would be left with learning 1 or random reorganization.

Is Trial and Error behavior different from control behavior?

bob

···
8

[Martin Taylor 2012.11.29.14.44]

[From Richard Pfau (2012.11.29 14:07 Nepal)]
...

From a PCT perspective, might the Law of Effect be restated as follows:

"Behaviors that reduce or eliminate error signals are more likely to occur again in similar situations. Conversely, behaviors that produce, increase, or fail to eliminate error signals are less likely to occur again in similar situations."

I ask if this restatement is OK, since I may include it in something that I am writing.

both of the rest of whatever you are writing and of the intended audience. If the concept of reorganization is explained either before or immediately following the quote, it is probably mostly OK -- but see below for why "mostly".

The basic principle of reorganization a la B:CP is that if error exists and persists or increases, then something about the structure below the failing control is likely to change. That is easily stated as "behaviors that produce, increase, or fail to eliminate error signals are less likely to occur again in similar situations". The converse, however is not true. Reorganization that changes the behaviour in a particular situation is not eliminated when the error is reduced. Its likelihood of happening in any given time interval is merely reduced. In the past, we have called the result the "winter leaf" phenomenon. Random winds move leaves in open spaces until they arrive at a more sheltered location from which they may occasionally be moved but are likely to stay for a while before that move happens. Leaves therefore pile up in drifts, just as though there was some agency trying to collect them into drifts.

If the B:CP principle of reorganization is even approximately correct, behaviours that reduce or eliminate errors are less likely to be changed than are those that don't. But they do not become more likely to happen than they were.

I somewhat agree with Kennaway's objection if "behaviour" and "situation" are defined by an external observer such as an experimenter, since the experimenter may not be correct about what the subject is perceiving nor about what aspect of what the subject is perceiving is actually being controlled. My agreement is only "somewhat", however, because the experimenter watching Fred's cat will consistently see the lever being pressed, but will (as Kennaway says) inconsistently see different muscular actions. The experimenter will reasonably see "lever pressing" as a behaviour, though what perception that behaviour is used to control may remain a mystery. Is "to perceive myself not in a cage" the main idea? Would a non-hungry cat not want to get out of the cage? There are opportunities for conducting The Test For The Controlled Variable, but most experimenters neither know nor care about the question.

If your readers are expected to understand the reorganization concept and the idea of behaviours at different levels of control, then your restatement is close to being OK in my eyes. I might suggest rewording the first part, though, along the lines of: "Behaviors that reduce or eliminate error signals tend to be retained for use in subsequent similar situations."

Martin

···

From my viewpoint, whether this statement is OK depends on the context,

9

[From Rick Marken (2012.11.29.2100)]

Richard Pfau (2012.11.29 14:07 Nepal)

RM: Are you really in Nepal Richard?!? If so, how cool (or maybe
"cold" is more appropriate;-)

RP: From a PCT perspective, might the Law of Effect be restated as follows:

"Behaviors that reduce or eliminate error signals are more likely to occur
again in similar situations. Conversely, behaviors that produce, increase,
or fail to eliminate error signals are less likely to occur again in similar
situations."

RM: How about "From a PCT perspective, the Law of Effect has been
repealed. The same behavior (output action) will rarely produce the
same effect (consequence); In order to produce desired effects --
effects that reduce error -- organisms must learn how to vary their
behaviors (output action) appropriately to produce the desired effect
while compensating for unpredictable and usually undetectable
disturbances".

The Law of Effect seems to hold when "behavior" is defined informally
and qualitatively. So it looks like the cat repeats the same "lever
pull" behavior to get out of the cage. But the term "behavior"
conflates means and ends and ignores variations in both. The lever
pull is an end result produced by varying means -- muscle forces and
movements -- depending on the cat's orientation relative to the lever.
The lever pull itself is a controlled result produced by varying these
varying means. It looks like a behavior (lever pull) is selected by
reward (effect) but it's actually the cat that has learned to select
(control for) the lever pull in order to control for the higher level
perception of the reward (effect). So from a PCT perspective I would
say that the Law of Effect has been repealed and replaced with the Law
of Control.

Best

Rick

···

--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

10

[From Erling Jorgensen (2012.11.30.1210 EST)]

Richard Pfau (2012.11.29 14:07 Nepal)

Hello, Richard. I, too, am curious about that "Nepal" designation.

From a PCT perspective, might the Law of Effect be restated as follows:

"Behaviors that reduce or eliminate error signals are more likely to
occur again in similar situations. Conversely, behaviors that produce,
increase, or fail to eliminate error signals are less likely to occur
again in similar situations."

You've heard from several on CSGNet that, from a PCT perspective, what is
called the Law of Effect is essentially an illusion. The illusion arises
from under-specifying what is more properly going on, that is, ignoring
differences that actually make a difference.

You can test out this so-called Law of Effect by producing the outcome of
standing on one foot. You should be able to reduce & keep the error
signals low in that situation. But that will hardly lead to recurring
behaviors. In fact, the behaviors typically change, in order to maintain
that recurring outcome.

That experiment, where stable outcomes come about from non-recurring
behaviors, is not an anomaly.

The one place where I share your intuition, which seemed to lead to
attempting to translate this 'Law'into PCT terms, is the correlation
with error signals. I agree that, from a PCT perspective, what is "good"
(by definition) is lower error & what is "bad" is higher error.

However, it is a more complicated matter whether "satisfying or pleasant"
feelings, versus "discomforting, annoying or unpleasant" feelings overlap
onto those foundational definitions of good or bad. I believe the
emotional components have to do with differential rates of change in
error for living control systems.

All the best,
Erling

11

[From Bill Powers (2012.11.30.1830 MST)]

Richard Pfau (2012.11.29 14:07 Nepal) –

From a PCT perspective, might the Law of Effect be restated as
follows:

“Behaviors that reduce or
eliminate error signals are more likely to occur again in similar
situations. Conversely, behaviors that produce, increase, or fail
to eliminate error signals are less likely to occur again in similar
situations.”

I ask if this restatement is OK, since I may include it in something that
I am writing.

I think we need a new word for behavior in this context. I tend to use
the term for the motor/glandular outputs that are directly caused by the
nervous system, which means that practically everything other people call
behavior I call consequences of behavior. The outputs that I call
behavior are not controlled variables in themselves, though they are
being used as a means of controlling some of ther variable.
On the other hand, there is the case in which one variable is controlled
because of its effect on another variable to be controlled. Example is
the position of a lever in a Skinner box, which is controlled as a means
of controlling the delivery of something the organism wants -=- a piece
of food dropping into the food dish. Neither the depression of the lever
nor the arrival of the food is a behavior because both are consequences
of the actual behavior – limb, body, and head movements.
To complicate matters further, the basic premise in the quotation is
based on the false assumption that to reproduce a given result of
behavior (like reducing an error signal), it is necessary to repeat the
behavior that occurred the last time that result occurred. If by behavior
you mean the motor acts, the most likely result of repeating the behavior
is to get a different result. This is because the environmental
details are always changing, largely owing to new disturbances or changes
in disturbances that were present the last time. Most often, the motor
actions must be varied in order to get the same result again.

This last problem is hard to get across because control is a process that
entails varying, not repeating, an action, with the result of repeating,
not varying the state of a controlled variable. It’s hard to find a way
of saying correctly that there is something that repeats when the same
result repeats. What is there, other than the result, that
repeats?

Best,

Bill P.

12

[From: Richard Pfau (2012.12.2. 2145 Nepal)

Thanks to all who provided their insights concerning the draft "Law of
Effect" restated.

As Rick Marken suggests, from a PCT perspective, the Law of Effect
seems to be repealed [ref Rick Marken (2012.11.29.2100)].

I thought we were on a roll when Martin Taylor (2012.11.29.14.44)
appeared to agree with the second sentence of the revised Law and suggested a change to the
first sentence, which, when combined with ideas from others seemed to
lead to the following "Revised Law of (Controlled) Effect":

"Body movements leading to consequences that reduce or
eliminate error signals are likely to occur again in situations
similarly perceived by an organism. Conversely, body movements that
produce, increase, or fail to eliminate error signals are less likely
to occur again in similar situations."

Does this seem a bit more on target,(especially when combined with
explanations of reorganization and other aspects of PCT, and
understanding that other body movements at other times can also lead to
similar consequences and subsequent reorganization for potential use if
another learned bodily movement doesn't work due to the effect of
disturbances) or does the statement still
seem so inconsistent with PCT as to be not worth stating?

I ask, since it seems that if PCT is to be accepted as a new paradigm
that replaces the old, bridges are needed to help those of the old
understanding move to the new. A maxim of education is to start where
the learner is and build upon his/her present knowledge. And so,
starting with a something like "The Law of Effect" that is familiar to
many who have studied psychology, and restating it in a way that helps
lead to understanding some basic PCT concepts, seems worth trying. But
then again, maybe this particular "Law" isn't worth pointing out as one
such possible bridge. In any case, it seems that if we can't start from
where people are now, and build upon at least some of their present understandings
to create insights into what PCT is all about, then PCT's entry into mainstream psychology
will take a very long time to occur.

Be that as it may, the feedback received from you-all has been
stimulating, and I am trying to take it fully into consideration.

With Regards,
Richard Pfau
[From the Kathmandu Valley near Swayambhunath (a Buddhist Stupa on a
hill, known to some visitors as "the Monkey Temple"), near the Himalayan
Mountain Range, where, yes, it is a bit cold, but where the friendly,
smiling people, landscape, and historic sights more than make up for
the chilliness -- (however, I must admit that it does help to wear long
underwear, wool socks, a warm sweater, and a snug hat!)]