http://www.biorxiv.org/content/early/2016/06/13/058719.abstract?%3Fcollection=
Is Brembs on the right track?
[From MK (2016.06.14.1215 CET)]
The abstract to his older article "The importance of Being Active"
contains a quite interesting sentence.
"Abstract: The successful stimulus-response approach to the
organization of behavior has been the dominating paradigm for much of
the psychology and neuroscience of the 20th century. Martin Heisenberg
is a pioneer in championing the idea that all brains, even
comparatively simple ones such as those of insects, instead operate
according to output-input principles. Since the 1970s, his research
produces evidence that the fruit fly, Drosophila melanogaster, is
capable of spontaneous behavioral activity, and that the flies use it
to control sensory input (i.e., operant behavior)."
···
----
Brembs, 2008: The Importance of Being Active
J. Neurogenetics, 2008, 1 7, iFirst article
DOI: 10.1080/01677060802471643
http://bjoern.brembs.net/wp/?wpdmdl=700
----
M
[From Rick Marken (2016.06.14.0930]
···
On Tue, Jun 14, 2016 at 2:15 AM, Alex Gomez-Marin agomezmarin@gmail.com wrote:
http://www.biorxiv.org/content/early/2016/06/13/058719.abstract?%3Fcollection=
AG: Is Brembs on the right track?
RM: The last sentence of the abstract that MK (2016.06.14.1215 CET) quotes sure makes it sound like he’s on the right track.
“Drosophila melanogaster, is capable of spontaneous behavioral activity, and that the flies use it to control sensory input (i.e., operant behavior).”
RM: But there a signs that he might end up going off the rails. For example, it’s not “spontaneous behavioral activity” that is used to control sensory input; it’s behavioral output variations that are quite “non-spontaneous” inasmuch as they must be continuously varying in precisely the right way so as to protect controlled sensory inputs from environmental variations (in disturbances and feedback functions) that would otherwise cause these sensory inputs to move away from their reference states.
RM: The best way to see whether Brembs is on the right track is to tell him about what should be his “destination” (PCT) and see what he does.
Best
Rick
–
Richard S. Marken
Author, with Timothy A. Carey, of Controlling People: The Paradoxical Nature of Being Human.
[From Bruce Abbott (2016.06.16.1000 EDT)]
Rick Marken (2016.06.14.0930 –
http://www.biorxiv.org/content/early/2016/06/13/058719.abstract?%3Fcollection=
AG: Is Brembs on the right track?
RM: The last sentence of the abstract that MK (2016.06.14.1215 CET) quotes sure makes it sound like he’s on the right track.
“Drosophila melanogaster, is capable of spontaneous behavioral activity, and that the flies use it to control sensory input (i.e., operant behavior).”
RM: But there a signs that he might end up going off the rails. For example, it’s not “spontaneous behavioral activity” that is used to control sensory input; it’s behavioral output variations that are quite “non-spontaneous” inasmuch as they must be continuously varying in precisely the right way so as to protect controlled sensory inputs from environmental variations (in disturbances and feedback functions) that would otherwise cause these sensory inputs to move away from their reference states.
Brembs’ “spontaneous activityâ€? is activity that arises from within the organism without being elicited in S-R fashion by an external stimulus. Here Brembs explicitly recognizes that “flies use such activity to control their sensory input.â€? When you assert that it’s not spontaneous behavioral activity that is used to control sensory input, you apply your own definition of “spontaneous activityâ€? –“ something more like “randomly generated activity.â€? Brembs could not have meant that, because such activity would control nothing.  But there is a way in which even “randomly generated activityâ€? could apply here. Organisms attempting to establish control over some perception typically try a number of actions until they hit upon one that works. In the operant conditioning literature the process is called “shaping.â€? In PCT it’s called “reorganization.â€?
Bruce
···
On Tue, Jun 14, 2016 at 2:15 AM, Alex Gomez-Marin agomezmarin@gmail.com wrote:
[From Rick Marken (2016.06.16.0825)]
···
Bruce Abbott (2016.06.16.1000 EDT)
Â
BA: Organisms attempting to establish control over some perception typically try a number of actions until they hit upon one that works. In the operant conditioning literature the process is called âshaping.â? In PCT itâs called âreorganization.â?
RM: Shaping and reorganization are two entirely different things. Shaping is done by a control system external to the control system whose behavior is being controlled (shaped); reorganization is done by a control system – the reorganization system – that is internal to the control system whose behavior is being controlled (reorganized).
BestÂ
Rick
–
Richard S. MarkenÂ
Author, with Timothy A. Carey, of  Controlling People: The Paradoxical Nature of Being Human.Â
[From Bruce Abbott (2016.16.1155 EDT)]
Rick Marken (2016.06.16.0825)
Bruce Abbott (2016.06.16.1000 EDT)
BA: Organisms attempting to establish control over some perception typically try a number of actions until they hit upon one that works. In the operant conditioning literature the process is called “shaping.� In PCT it’s called “reorganization.�
RM: Shaping and reorganization are two entirely different things. Shaping is done by a control system external to the control system whose behavior is being controlled (shaped); reorganization is done by a control system – the reorganization system – that is internal to the control system whose behavior is being controlled (reorganized).
BA: As in the previous exchange, the disagreement comes down to a difference in definition.  By “shaping� you mean a technique used by a trainer, whereas I mean a process that can be put into play by a trainer but which also occurs naturally. In the natural process, a variant action that controls better tends to replace previous variants.  The distinction I make is similar to the distinction Darwin made between “artificial� selection (a breeder does the selecting) and “natural� selection (differential survival and reproductive fitness).
Bruce
I think Bruce has put it very well: Brembs obviously has the intuition that behavior is not and cannot be stimulus-response, yet he has gone all the way to the other extreme to free behavior from the tyranny of reflex chains and proclaim that it as absolutely spontaneous, nearly indetermined, most likely missing that the stimulus-response, and its action-outcome “upgrade”, can both meet and gain depth in the synthesis offered by control loops.
It is ironic that perhaps the two most studied behavioral situations in neuroscience are either super-constrained tasks with strong stimuli and allowing little explicit control from the animal (ie. learning in a box) or un-constrained situations with poverty of stimulus (not necessarily of perception) where the animal is trying to figure out how to control something at all and eventually starting a process of reorganization (ie. exploring/foraging in an empty arena).Â
Borges’ tale of The two kings and the two labyrinths: one placed in a super complex human-made maze, the other dropped and lost in the desert…
···
On Thu, Jun 16, 2016 at 5:56 PM, Bruce Abbott bbabbott@frontier.com wrote:
[From Bruce Abbott (2016.16.1155 EDT)]
Â
Rick Marken (2016.06.16.0825)
Â
Bruce Abbott (2016.06.16.1000 EDT)
 BA: Organisms attempting to establish control over some perception typically try a number of actions until they hit upon one that works. In the operant conditioning literature the process is called “shaping.� In PCT it’s called “reorganization.�
Â
RM: Shaping and reorganization are two entirely different things. Shaping is done by a control system external to the control system whose behavior is being controlled (shaped); reorganization is done by a control system – the reorganization system – that is internal to the control system whose behavior is being controlled (reorganized).
Â
BA: As in the previous exchange, the disagreement comes down to a difference in definition. By “shaping� you mean a technique used by a trainer, whereas I mean a process that can be put into play by a trainer but which also occurs naturally. In the natural process, a variant action that controls better tends to replace previous variants. The distinction I make is similar to the distinction Darwin made between “artificial� selection (a breeder does the selecting) and “natural� selection (differential survival and reproductive fitness).
Â
Bruce
Â
[From Rick Marken (2016.06.16.1015)]
···
Bruce Abbott (2016.16.1155 EDT)
Â
Rick Marken (2016.06.16.0825)
Â
RM: Shaping and reorganization are two entirely different things. Shaping is done by a control system external to the control system whose behavior is being controlled (shaped); reorganization is done by a control system – the reorganization system – that is internal to the control system whose behavior is being controlled (reorganized).
Â
BA: As in the previous exchange, the disagreement comes down to a difference in definition. By “shapingâ€? you mean a technique used by a trainer, whereas I mean a process that can be put into play by a trainer but which also occurs naturally.Â
RM: That’s the mistake Skinner made. He assumed that the environment (nature) shapes behavior just like he did. But, unlike Skinner, the environment is not a control system; it doesn’t shape behavior because it has no goals (references) for how an organism should behave so it doesn’t act (as Skinner does) by “reinforcing” successive approximations to the desired (goal) behavior. This is all explained in Chapter 6 of “Controlling People”; get your copy today!!
BA: In the natural process, a variant action that controls better tends to replace previous variants. The distinction I make is similar to the distinction Darwin made between “artificial� selection (a breeder does the selecting) and “natural� selection (differential survival and reproductive fitness).
RM: That is a pretty good description of reorganization, though you have to substitute “control organization” for “action”. The result of reorganization is a control system that is better organized to control (vary its actions as needed to keep the controlled variable in the specified reference state). The environment (nature) doesn’t select this control organization; the reorganization system inside the organism selects control organizations that are good at protecting controlled variables from the unpredictable (and often undetectable) effects of the environment.Â
Best
Rick
–
Richard S. MarkenÂ
Author, with Timothy A. Carey, of  Controlling People: The Paradoxical Nature of Being Human.Â
[From Bruce Abbott (2016.06.16.1950 EDT)]
Rick Marken (2016.06.16.1015) –
Bruce Abbott (2016.16.1155 EDT)
Rick Marken (2016.06.16.0825)
RM: Shaping and reorganization are two entirely different things. Shaping is done by a control system external to the control system whose behavior is being controlled (shaped); reorganization is done by a control system – the reorganization system – that is internal to the control system whose behavior is being controlled (reorganized).
BA: As in the previous exchange, the disagreement comes down to a difference in definition. By “shaping� you mean a technique used by a trainer, whereas I mean a process that can be put into play by a trainer but which also occurs naturally.
RM: That’s the mistake Skinner made. He assumed that the environment (nature) shapes behavior just like he did. But, unlike Skinner, the environment is not a control system; it doesn’t shape behavior because it has no goals (references) for how an organism should behave so it doesn’t act (as Skinner does) by “reinforcing” successive approximations to the desired (goal) behavior. This is all explained in Chapter 6 of “Controlling People”; get your copy today!!
BA: Skinner did not make that mistake. Shaping as a consequence of environmental contingencies does not involve goals on the part of the environment, and Skinner did not assume that it does (see below).
BA: In the natural process, a variant action that controls better tends to replace previous variants. The distinction I make is similar to the distinction Darwin made between “artificial� selection (a breeder does the selecting) and “natural� selection (differential survival and reproductive fitness).
RM: That is a pretty good description of reorganization, though you have to substitute “control organization” for “action”.
BA: Yes, except here in this discussion the focus is on a particular aspect of reorganization, variation in actions from which improved control may emerge.
RM: The result of reorganization is a control system that is better organized to control (vary its actions as needed to keep the controlled variable in the specified reference state). The environment (nature) doesn’t select this control organization; the reorganization system inside the organism selects control organizations that are good at protecting controlled variables from the unpredictable (and often undetectable) effects of the environment.
BA: The role of selection in the Skinnerian system is played by events called reinforcers and punishers. Whether an event will function as either for a given individual, at a given time, under given conditions, depends on the internal organization and physiological state of that individual. The environment shapes behavior, not because it has goals for how an organism should behave, but because it presents natural contingencies between an organism’s behaviors and their consequences. As behavior varies, those actions that produce reinforcers become more probable, relative to other possible actions. Those actions that produce punishers become less probable, relative to other possible actions. Skinner termed this selective process “selection by consequences.� It is directly analogous to Darwin’s natural selection. As in Darwin’s process, selection by consequences is not done by a control system and does not involve any goals for how the individual should behave.
BA: If we equate reinforcers with inputs that reduce error in a controlled perception and punishers with inputs that increase such error, then we can understand natural shaping in terms of reorganization. Of the various candidate actions that reorganization provides, those that reduce error (in Skinner’s terms, produce reinforcers) will tend to be retained; those that increase error (in Skinner’s terms, produce punishers) or fail to reduce it will tend to be abandoned. The result is selection toward more and more effective forms of action – shaping.
Bruce
Rick and Bruce:
Wouldn't 'shaping' just be the environmental requirements that must
be met to achieve control? That is, they could be no more than just
the physical laws of mechanics, the actions of another control
system (living or engineered), or both.
If I'm not missing the finer points that you two are talking about,
the above explains how the environment results in changes (shaping)
caused by reorganization irrespective of whether there is any
external control system involved or not.
Best,
Bill
···
On 06/16/2016 09:56 AM, Bruce Abbott
wrote:
[From Bruce Abbott (2016.16.1155 EDT)]
Â
Rick Marken (2016.06.16.0825)
Â
Bruce Abbott (2016.06.16.1000EDT)
 BA: Organisms attempting toestablish control over some perception typically try
a number of actions until they hit upon one that
works. In the operant conditioning literature the
process is called “shaping.� In PCT it’s called
“reorganization.�
Â
RM: Shaping and reorganization aretwo entirely different things. Shaping is done by a
control system external to the control system whose
behavior is being controlled (shaped); reorganization
is done by a control system – the reorganization
system – that is internal to the control system whose
behavior is being controlled (reorganized).
Â
BA: As in the previous exchange, the disagreement comesdown to a difference in definition.  By “shaping� you mean
a technique used by a trainer, whereas I mean a process
that can be put into play by a trainer but which also
occurs naturally. In the natural process, a variant
action that controls better tends to replace previous
variants. Â The distinction I make is similar to the
distinction Darwin made between “artificial� selection (a
breeder does the selecting) and “natural� selection
(differential survival and reproductive fitness).
Â
Bruce
Â
[From Rick Marken (2016.06.17.1030)]
Bruce Abbott (2016.06.16.1950 EDT)]
Â
RM: That's the mistake Skinner made. He assumed that the environment (nature) shapes behavior just like he did. But, unlike Skinner, the environment is not a control system; it doesn't shape behavior because it has no goals (references) for how an organism should behave so it doesn't act (as Skinner does) by "reinforcing" successive approximations to the desired (goal) behavior. This is all explained in Chapter 6 of "Controlling People"; get your copy today!!
Â
BA: Skinner did not make that mistake. Shaping as a consequence of environmental contingencies does not involve goals on the part of the environment, and Skinner did not assume that it does (see below).
RM: Of course he didn't explicitly assume that it involves goals on the part of the environment. He didn't even understand that it required goals on his part either. Skinner didn't know (or want to know) that what he was doing when he was shaping behavior was doing something purposeful. He didn't even know what purpose was (spoiler alert: it's controlling). That was Skinner's great failure; we was a master at controlling behavior and he never understood the nature of what he was doing and how that differed from causing things to happen.
BA: If we equate reinforcers with inputs that reduce error in a controlled perception and punishers with inputs that increase such error, then we can understand natural shaping in terms of reorganization. Of the various candidate actions that reorganization provides, those that reduce error (in Skinner’s terms, produce reinforcers) will tend to be retained; those that increase error (in Skinner’s terms, produce punishers) or fail to reduce it will tend to be abandoned. The result is selection toward more and more effective forms of action – shaping.
RM: Whatt is a reinforcer and what is a punisher is defined by the organism (PCT says the are defined by reference signals in the organism). So the environmental events that correspond to reinforcers and punishers have no say in whether they will be a reinforcer or a punisher. So a piece of cow dung can't determine whether it will be reinforcing or punishing; it's just sitting out there being cow dung. But it can give the illusion of selecting the behavior. So the same piece of cow dung will appear to select approach behavior in flies and avoidance behavior in hikers. This is the illusion that Skinner (and virtually all "scientific psychologists) fell for.Â
RM: In order to understand the difference between the actual shaping that Skinner did and the apparent shaping that the environment does you have to understand what controlling IS. Controlling involves bringing a variable aspect of the environment (the controlled variable) to a reference state while acting to protect it from disturbances. This is what Skinner was doing when he shaped behavior. The controlled variable was some aspect of an organism's behavior, like how often a pigeon pecked a key. Disturbances to this variable are variations in the behavior, such as variations in how often the pigeon actually pecks the key. The actions that compensate for these disturbances are the giving and withholding of "reinforcers" -- things that the organism is known to want. The shaper brings the controlled variable to the reference state by providing reinforcements only when the organism produces behaviors that more closely approximate the reference behavior; "reinforcement of successive approximations". So the key pecking behavior, which is initially at  0 pecks/sec is carefully brought to a reference state > 0 pecks/sec and maintained in that state by appropriate provision and withholding of reinforcers.Â
RM: So shaping is a control process; it can only be done by a control system -- one that can perceive and have a reference for the state of the behavioral variable and can act, as necessary, (by giving and withholding reinforcers) to bring that variable to the reference state and keep it there, protected from disturbances (changes in what the organism does). This is clearly not what is going on in the natural environment. The same "reinforcers" may be present in the environment with the same organism (pigeon) but those reinforcers don't have a reference for what the organism should do (they have no reference for the rate at which the pigeon pecks the key, for example). So they won't provide or withhold themselves depending on what the organism does. They are not controlling the organism's behavior.Â
RM: So that's the difference between the actual shaping done by a controller (like Skinner) and the apparent shaping done by the environment. In one case their is specific, pre-selected result (the desired behavior of the organism) and in the other there is not. In one case an organism's behavior is being controlled by an outside agent; in the other it is not. Shaping is controlling. Skinner can control; the environment can't.Â
BestÂ
Rick
···
--
Richard S. MarkenÂ
Author, with Timothy A. Carey, of  <https://urldefense.proofpoint.com/v2/url?u=http-3A__www.amazon.com_Controlling-2DPeople-2DParadoxical-2DNature-2DBeing_dp_1922117641_ref-3Dsr-5F1-5F1-3Fs-3Dbooks-26ie-3DUTF8-26qid-3D1449541975-26sr-3D1-2D1&d=CwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=-dJBNItYEMOLt6aj_KjGi2LMO_Q8QB-ZzxIZIF8DGyQ&m=Ah5UIka1M4T4MZaYyTbgqrmSi38XCLojr30MTUUAbvg&s=d5eLXCutUvu-CgP9fJ-QiBHG2v1wN1HjeF2ysG2DV7o&e=>Controlling People: The Paradoxical Nature of Being Human.Â
[From Bruce Abbott (2016.20.1900 EDT)]
Rick Marken (2016.06.17.1030) –
Bruce Abbott (2016.06.16.1950 EDT)]
RM: That’s the mistake Skinner made. He assumed that the environment (nature) shapes behavior just like he did. But, unlike Skinner, the environment is not a control system; it doesn’t shape behavior because it has no goals (references) for how an organism should behave so it doesn’t act (as Skinner does) by “reinforcing” successive approximations to the desired (goal) behavior. This is all explained in Chapter 6 of “Controlling People”; get your copy today!!
BA: Skinner did not make that mistake. Shaping as a consequence of environmental contingencies does not involve goals on the part of the environment, and Skinner did not assume that it does (see below).
RM: Of course he didn’t explicitly assume that it involves goals on the part of the environment. He didn’t even understand that it required goals on his part either. Skinner didn’t know (or want to know) that what he was doing when he was shaping behavior was doing something purposeful. He didn’t even know what purpose was (spoiler alert: it’s controlling). That was Skinner’s great failure; we was a master at controlling behavior and he never understood the nature of what he was doing and how that differed from causing things to happen.
BA: I agree with the first sentence above, and would add to it, “nor did he explicitly assume them.â€? I’m quite sure Skinner understood that the shaping procedure required goals on his part, or for that matter on the part of anyone engaging in the process. It’s obvious – you have to know what you are trying to accomplish. As for not understanding the nature of what he was doing (control), he understood it well enough to develop the prototype of the world’s first “smart bombâ€? – one that would apply corrective actiions to its fins in order to steer the bomb toward its target.Â
BA: But enough of Skinner – being deceased he’s no longer availlable to clarify his views on this matter.
BA: If we equate reinforcers with inputs that reduce error in a controlled perception and punishers with inputs that increase such error, then we can understand natural shaping in terms of reorganization. Of the various candidate actions that reorganization provides, those that reduce error (in Skinner’s terms, produce reinforcers) will tend to be retained; those that increase error (in Skinner’s terms, produce punishers) or fail to reduce it will tend to be abandoned. The result is selection toward more and more effective forms of action – shapinng
RM: What is a reinforcer and what is a punisher is defined by the organism (PCT says the are defined by reference signals in the organism). So the environmental events that correspond to reinforcers and punishers have no say in whether they will be a reinforcer or a punisher. So a piece of cow dung can’t determine whether it will be reinforcing or punishing; it’s just sitting out there being cow dung. But it can give the illusion of selecting the behavior. So the same piece of cow dung will appear to select approach behavior in flies and avoidance behavior in hikers. This is the illusion that Skinner (and virtually all "scientific psychologists) fell for.
BA:Â I can come up with no better reply than to past in the paragraph from my post that you omitted, which speaks directly to this issue:
BA: The role of selection in the Skinnerian system is played by events called reinforcers and punishers. Whether an event will function as either for a given individual, at a given time, under given conditions, depends on the internal organization and physiological state of that individual. The environment shapes behavior, not because it has goals for how an organism should behave, but because it presents natural contingencies between an organism’s behaviors and their consequences. As behavior varies, those actions that produce reinforcers become more probable, relative to other possible actions. Those actions that produce punishers become less probable, relative to other possible actions. Skinner termed this selective process “selection by consequences.� It is directly analogous to Darwin’s natural selection. As in Darwin’s process, selection by consequences is not done by a control system and does not involve any goals for how the individual should behave.
BA: We agree: “A piece of cow dung can’t determine whether it will be reinforcing or punishing.� The above paragraph explains what is required. What is delivered as a consequence of a behavioral act will function as a reinforcer for the same reason it will function to reduce error in a controlled variable. In both cases it depends on the relationship between the properties of the reinforcers and the internal organization and state of the organism.
RM: In order to understand the difference between the actual shaping that Skinner did and the apparent shaping that the environment does you have to understand what controlling IS. Controlling involves bringing a variable aspect of the environment (the controlled variable) to a reference state while acting to protect it from disturbances. This is what Skinner was doing when he shaped behavior. The controlled variable was some aspect of an organism’s behavior, like how often a pigeon pecked a key. Disturbances to this variable are variations in the behavior, such as variations in how often the pigeon actually pecks the key. The actions that compensate for these disturbances are the giving and withholding of “reinforcers” – things that the organism is known to want. The shaper brings the controlled variable to the reference state by providing reinforcements only when the organism produces behaviors that more closely approximate the reference behavior; “reinforcement of successive approximations”. So the key pecking behavior, which is initially at 0 pecks/sec is carefully brought to a reference state > 0 pecks/sec and maintained in that state by appropriate provision and withholding of reinforcers.
BA: I don’t know why you are explaining this to me as if I had not already said as much myself:
BA: If we equate reinforcers with inputs that reduce error in a controlled perception and punishers with inputs that increase such error, then we can understand natural shaping in terms of reorganization. Of the various candidate actions that reorganization provides, those that reduce error (in Skinner’s terms, produce reinforcers) will tend to be retained; those that increase error (in Skinner’s terms, produce punishers) or fail to reduce it will tend to be abandoned. The result is selection toward more and more effective forms of action – shaping.
RM: So shaping is a control process; it can only be done by a control system – one that can perceive and have a reference for the state of the behavioral variable and can act, as necessary, (by giving and withholding reinforcers) to bring that variable to the reference state and keep it there, protected from disturbances (changes in what the organism does). This is clearly not what is going on in the natural environment. The same “reinforcers” may be present in the environment with the same organism (pigeon) but those reinforcers don’t have a reference for what the organism should do (they have no reference for the rate at which the pigeon pecks the key, for example). So they won’t provide or withhold themselves depending on what the organism does. They are not controlling the organism’s behavior.
BA:  You want us to believe that by the phrase, “selecting by consequences,� Skinner meant that the consequences are some kind of agent actively selecting behavior. He doesn’t. The phrase means the following: Whether the organism will become more (or less) likely to produce a given behavior depends on what the consequences of that behavior are for the organism. The internal organization and state of the organism at the time that a given consequence occurs determines its effect on that organism’s behavior.
BA: What is that internal organization? What is the current internal state of the organism? Skinner did not wish to speculate. (He wrote that was a job best left to the physiologist.)  For some cases it seems fairly obvious, as when food serves as a reinforcer after a short period of food deprivation, but it’s not so obvious in other cases, so he was content to simply observe the conditions under which a given consequence would in fact reinforce given act.
BA: In natural shaping, what particular acts will produce a given reinforcer, and how efficiently they will do so, depends on the environment – what we in PCT call the feedback function. If the organnism is “spontaneously active,â€? those variant activities that chance to produce the reinforcer will increase in probability relative to other activities, and those variants of that effective act that do so more effectively will be selected over those that are less efficient. This isn’t some theoretical notion – you can observe the proceess in action. From the observer’s point of view, the consequences of the organism’s activities drive the selective process, determining which activities will be promoted and gradually shaped into a relatively efficient means of obtaining those consequences. That is “selection by consequences.â€? Nothing here either asserts or implies that those consequences possess some magical ability to do the selecting – that is done by the organism (by its internal mechanisms) on the basis of the consequences.
BA: What PCT offers is an explanation for the effectiveness of these consequences in bringing about changes in behavior. The language of operant conditioning describes observable, repeatable relationships, but it only describes them. PCT describes the mechanism that enforces those observed relationships.
BA: In the past, the differences between the Skinnerian approach and PCT have been compared on this list to the rivalry between the old Ptolemaic, earth-centered view of the heavens and the Copernican sun-centered view, but this is a poor analogy. A better one is the difference between Kepler’s laws of planetary motion and Newton’s laws of motion and gravity. Kepler’s laws described how the planets move; Newton’s laws explained why they move as Kepler described.
Bruce
[From Rick Marken (2016.06.21.1950)]
···
Bruce Abbott (2016.20.1900 EDT)
Â
RM: That’s the mistake Skinner made. He assumed that the environment (nature) shapes behavior just like he did. But, unlike Skinner, the environment is not a control system; it doesn’t shape behavior because it has no goals (references) for how an organism should behave so it doesn’t act (as Skinner does) by “reinforcing” successive approximations to the desired (goal) behavior. This is all explained in Chapter 6 of “Controlling People”; get your copy today!!
Â
BA: Skinner did not make that mistake. Shaping as a consequence of environmental contingencies does not involve goals on the part of the environment, and Skinner did not assume that it does (see below).
Â
RM: I didn’t say that Skinner’s mistake was to think that environmental shaping involved goals like it does in his own shaping; his mistake was to think that shaping doesn’t involve goals at all whether the shaping is done by the environment or by a person (like himself).
RM: Of course he didn’t explicitly assume that [shaping] involves goals on the part of the environment. He didn’t even understand that it required goals on his part either.
Â
BA: I agree with the first sentence above, and would add to it, “nor did he explicitly assume them.� I’m quite sure Skinner understood that the shaping procedure required goals on his part, or for that matter on the part of anyone engaging in the process.
RM:  Skinner may have “known” that he was working to achieve goals when he was shaping behavior – that is, he may have known that shaping seemed to be purposeful – but he did everything he could to prove that purpose was actually an illusion. Indeed, he argued that his own behavior  – including the behavior of shaping behavior – was itself shaped by the environment. According to Skinner his own shaping behavior was shaped by the behavior of the animal he was shaping; he called it “reciprocal control”, remember.Â
 BA: It’s obvious – you have to know what you are trying to accoomplish.Â
RM: It’s obvious to you but it was not obvious to Skinner or any of his followers.Â
BA: As for not understanding the nature of what he was doing (control), he understood it well enough to develop the prototype of the world’s first “smart bombâ€? – one that would apply corrrective actions to its fins in order to steer the bomb toward its target.Â
RM: He didn’t need to understand that his own behavior was a process of control in order to train a pigeon to  keep a cursor on target by pecking. Indeed, thousands of people – psychologists, animal trainers and control engineers – were able to produce systems that control (like the pigeon in the missile controlling the cursor) without knowing that they did it by controlling those systems. Indeed, Bill Powers was the first one to understand that the behavior of living organisms IS control. So Skinner was not alone in not knowing that he was controlling. And there are still only a few people in the world who know that behavior is control. But Tim and I are trying to up that number with our book “Controlling People”. You should get a copy, or two;-)Â
RM: In order to understand the difference between the actual shaping that Skinner did and the apparent shaping that the environment does you have to understand what controlling IS. Â
Â
BA: I don’t know why you are explaining this to me as if I had not already said as much myself:
Â
BA: If we equate reinforcers with inputs that reduce error in a controlled perception and punishers with inputs that increase such error, then we can understand natural shaping in terms of reorganization.
RM: The is no such thing as an input that reduces error? It’s the actions that affect the state of an input that reduce error (if the feedback in the loop is negative). When you say a reinforcer is an input that reduces error you are verbally turning inputs into variables that cause what the system does – the input-output model of behavior. But in a control system, the inputs are controlled by the system; they don’t cause (or control or select) the behavior of the system In fact, reinforcers, like food pellets, are not inputs at all but independent events that are disturbances to controlled variables. Organisms act to prevent these disturbances from having an effect on the controlled variable. So pigeons peck keys to control their perception of getting fed by preventing the food pellets from disturbing this variable by not coming often enough. The organism is controlling its input (the state of the controlled variable); the input is not controlling (“shaping”) the organism.
BA:  You want us to believe that by the phrase, “selecting by consequences,â€? Skinner meant that the consequences are some kind of agent actively selecting behavior. He doesn’t.Â
RM: I know. And this process is not the same as the “shaping” carried out by Skinner, which involves selection of consequences (in the case of shaping, selection of a consequence of giving and withholding reinforcers, like a pigeon playing ping pong).Â
Â
BA: In natural shaping, what particular acts will produce a given reinforcer, and how efficiently they will do so, depends on the environment – what we in PCT call the feedback ffunction. If the organism is “spontaneously active,â€? those variant activities that chance to produce the reinforcer will increase in probability relative to other activities, and those variants of that effective act that do so more effectively will be selected over those that are less efficient.Â
 RM: Was it natural shaping or shaping done by some amazing shaper that reinforced Beethoven’s “acts” (the Sonatas, quartets, Symphones, etc),  Shakespeare’s plays, Newton’s Principia, Joyce’s Ulysses?Â
BA: This isn’t some theoretical notion – you can observe the process in acction.
RM: Yes ,you can see it. It’s called “the behavioral illusion”; the illusion that inputs select outputs (See Powers’ reply to operant conditioners Baum and Reese starting on p. 83 of LCS I. I imagine you have already read it and dismissed it. But I think Powers made some great points, Â especially when he said: “If control-system theory does indeed describe correctly the relationship between organisms and their environments, behaviorism has been in the grip of a powerful illusion since its conceptual foundations were laid”.
Â
BA: What PCT offers is an explanation for the effectiveness of these consequences in bringing about changes in behavior. The language of operant conditioning describes observable, repeatable relationships, but it only describes them. PCT describes the mechanism that enforces those observed relationships.
RM: Yes, it sure does. It shows that what appears to be going on – selection  BY consequences – is not what’s going on. What’s going on is selection OF consequences: control. It looks like consequences select actions but it’s actually the other way around. And the consequences that are selected are not reinforcers; they are functions of reinforcers called controlled variables.
Â
BA: In the past, the differences between the Skinnerian approach and PCT have been compared on this list to the rivalry between the old Ptolemaic, earth-centered view of the heavens and the Copernican sun-centered view, but this is a poor analogy. A better one is the difference between Kepler’s laws of planetary motion and Newton’s laws of motion and gravity. Kepler’s laws described how the planets move; Newton’s laws explained why they move as Kepler described.Â
RM: I think neither analogy fits, mainly because Skinner (and all other psychologists) failed to observe what turns out to be the most important fact about operant (or any) behavior; that some variable aspect of the environment (like rate of shock) is being maintained in a reference state. (See Henry Yin’s paper “Restoring Purpose in Behavior” (https://www.dropbox.com/s/jgbtvvfu0zpz65j/Yin.pdf?dl=0), particularly section 6 and Figure 5. Now THAT is the PCT approach to dealing with operant behavior.Â
BestÂ
Rick
Richard S. MarkenÂ
Author, with Timothy A. Carey, of  Controlling People: The Paradoxical Nature of Being Human.Â
[From Bruce Abbott (2016.06.22.1955 EST)]
Rick Marken (2016.06.21.1950) –
Bruce Abbott (2016.20.1900 EDT)
RM: That’s the mistake Skinner made. He assumed that the environment (nature) shapes behavior just like he did. But, unlike Skinner, the environment is not a control system; it doesn’t shape behavior because it has no goals (references) for how an organism should behave so it doesn’t act (as Skinner does) by “reinforcing” successive approximations to the desired (goal) behavior. This is all explained in Chapter 6 of “Controlling People”; get your copy today!!
BA: Skinner did not make that mistake. Shaping as a consequence of environmental contingencies does not involve goals on the part of the environment, and Skinner did not assume that it does (see below).
RM: I didn’t say that Skinner’s mistake was to think that environmental shaping involved goals like it does in his own shaping; his mistake was to think that shaping doesn’t involve goals at all whether the shaping is done by the environment or by a person (like himself).
RM: Of course he didn’t explicitly assume that [shaping] involves goals on the part of the environment. He didn’t even understand that it required goals on his part either.
BA: I agree with the first sentence above, and would add to it, “nor did he explicitly assume them.� I’m quite sure Skinner understood that the shaping procedure required goals on his part, or for that matter on the part of anyone engaging in the process.
RM: Skinner may have “known” that he was working to achieve goals when he was shaping behavior – that is, he may have known that shaping seemed to be purposeful – but he did everything he could to prove that purpose was actually an illusion. Indeed, he argued that his own behavior – including the behavior of shaping behavior – was itself shaped by the environment. According to Skinner his own shaping behavior was shaped by the behavior of the animal he was shaping; he called it “reciprocal control”, remember.
BA: Ah HA! Something we agree on! From Skinner’s point of view, shaping is analogous to evolution by natural selection, which results in organisms becoming better adapted to their environment even though purpose plays no part in the process. Indeed, Skinner would give no credit to himself for his accomplishments, contending instead that they were simply products of the environmental contingencies that had acted upon him over his lifetime. But this ignores the fact that behavior is organized in the service of control, and that what and how we control are products of both our innate natures and the environment within which our control systems must act.
BA: It’s obvious – you have to know what you are trying to accommplish.
RM: It’s obvious to you but it was not obvious to Skinner or any of his followers.
BA: As for not understanding the nature of what he was doing (control), he understood it well enough to develop the prototype of the world’s first “smart bombâ€? – one that would apply corrective acttions to its fins in order to steer the bomb toward its target.
RM: He didn’t need to understand that his own behavior was a process of control in order to train a pigeon to keep a cursor on target by pecking. Indeed, thousands of people – psychologists, animal trainers and control engineers – were able to produce systems that control (like the pigeon in the missile controlling the cursor) without knowing that they did it by controlling those systems. Indeed, Bill Powers was the first one to understand that the behavior of living organisms IS control. So Skinner was not alone in not knowing that he was controlling. And there are still only a few people in the world who know that behavior is control. But Tim and I are trying to up that number with our book “Controlling People”. You should get a copy, or two;-)
BA: People have been creating useful control systems for millennia without having a detailed understanding of control theory. Skinner’s “smart bomb� was to be used on enemy ships. An image of the target ship appeared on a screen and the mechanism Skinner designed changed the bomb’s fin angles so as to bring the image of the ship onto the center of the screen. There was a pair of simple feedback loops (for the horizontal and vertical fins) designed to reduce the error between the position of the image and the desired position. Skinner certainly understood that.
BA: The key element that Skinner introduced was to have the location of the target on the screen indicated electronically by where on the screen a pigeon pecked. Skinner trained his pigeons to peck at the target image. The difference between this location and screen center determined the fin angles, which steered the bomb so as to zero out this difference. The pigeons were even able to distinguish the aerial image of the intended target ship from that of other ships, or in other words, to discriminate target from similar non-targets.
BA: If we equate reinforcers with inputs that reduce error in a controlled perception and punishers with inputs that increase such error, then we can understand natural shaping in terms of reorganization.
RM: The is no such thing as an input that reduces error? It’s the actions that affect the state of an input that reduce error (if the feedback in the loop is negative). When you say a reinforcer is an input that reduces error you are verbally turning inputs into variables that cause what the system does – the input-output model of behavior.
BA: I have a hunger-control system whose reference is zero hunger. By definition, if I am hungry then there is error present in that system. If you input food to this system, hunger diminishes. Is that not “an input that reduces error?� Of course it is!
BA: The food may have been consumed because the error causes food to be eaten (the action). But it is not eating (the action) that reduces the error, it is the food. Also, the food will reduce the error even if directly injected into the stomach. In that case it serves as a disturbance to the hunger control system – a helpful one in this case since it will reduce the error. So the food can reduce error (if error is present) whether it enters the control system via the system’s own actions or via an independent agent.
RM: But in a control system, the inputs are controlled by the system; they don’t cause (or control or select) the behavior of the system In fact, reinforcers, like food pellets, are not inputs at all but independent events that are disturbances to controlled variables. Organisms act to prevent these disturbances from having an effect on the controlled variable. So pigeons peck keys to control their perception of getting fed by preventing the food pellets from disturbing this variable by not coming often enough. The organism is controlling its input (the state of the controlled variable); the input is not controlling (“shaping”) the organism.
BA: To analyze the situation correctly we have to distinguish carefully between two situations. One is the operation of an established control system. The other is reorganization. Reorganization can create new control systems, but most of the time it acts to modify existing systems. In this discussion I have been focusing on the process of shaping and have proposed that shaping involves the reorganization of the output mechanism. If the organism is experiencing error in a controlled perception, it will perform the actions dictated by its current organization, actions that in the past have succeeded in reducing the error to acceptable levels. If these now consistently fail, other actions will be tried as reorganization varies the output function. Those actions that succeed in restoring control will be retained. That is selection by consequences.
BA: In the Skinnerian system, actions are selected when their consequence is the production of reinforcers, or the prevention/removal of punishers. In PCT, actions are selected at random and retained to the extent they succeed in reducing error in a given control system. Either way, it’s selection by consequences.Â
BA: You want us to believe that by the phrase, “selecting by consequences,� Skinner meant that the consequences are some kind of agent actively selecting behavior. He doesn’t.
RM: I know. And this process is not the same as the “shaping” carried out by Skinner, which involves selection of consequences (in the case of shaping, selection of a consequence of giving and withholding reinforcers, like a pigeon playing ping pong).
BA: Wrong. It’s the same shaping process whether carried out by an external agent or as a consequence of natural environmental contingencies.
BA: In natural shaping, what particular acts will produce a given reinforcer, and how efficiently they will do so, depends on the environment – what we in PCT call the feedback function. If the organism is “spontaneously active,� those variant activities that chance to produce the reinforcer will increase in probability relative to other activities, and those variants of that effective act that do so more effectively will be selected over those that are less efficient.
RM: Was it natural shaping or shaping done by some amazing shaper that reinforced Beethoven’s “acts” (the Sonatas, quartets, Symphones, etc), Shakespeare’s plays, Newton’s Principia, Joyce’s Ulysses?
BA: Probably both. Given that selection by consequences is how reorganization works, one will be as hard pressed to explain these extraordinary performances from the PCT perspective as from the Skinnerian.
BA: This isn’t some theoretical notion – you can observe the pprocess in action.
RM: Yes ,you can see it. It’s called “the behavioral illusion”; the illusion that inputs select outputs (See Powers’ reply to operant conditioners Baum and Reese starting on p. 83 of LCS I. I imagine you have already read it and dismissed it. But I think Powers made some great points, especially when he said: “If control-system theory does indeed describe correctly the relationship between organisms and their environments, behaviorism has been in the grip of a powerful illusion since its conceptual foundations were laid”.
BA: Shaping has nothing to do with the behavioral illusion, i.e., mistaking the inverse of the feedback function (a property of the environment) for a property of the organism’s behavioral system. Shaping (natural or externally directed) determines which among various candidate output functions will be retained by a given control system.
BA: What PCT offers is an explanation for the effectiveness of these consequences in bringing about changes in behavior. The language of operant conditioning describes observable, repeatable relationships, but it only describes them. PCT describes the mechanism that enforces those observed relationships.
RM: Yes, it sure does. It shows that what appears to be going on – selection BY consequences – is not what’s going on. What’s going on is selection OF consequences: control. It looks like consequences select actions but it’s actually the other way around. And the consequences that are selected are not reinforcers; they are functions of reinforcers called controlled variables.
BA: So, according to you (your words), in “selection OF consequences,� actions select consequences. And those consequences are controlled variables. I don’t recall any mechanisms in PCT by which an output of a control system selects controlled variables. Please explain!
BA: In the past, the differences between the Skinnerian approach and PCT have been compared on this list to the rivalry between the old Ptolemaic, earth-centered view of the heavens and the Copernican sun-centered view, but this is a poor analogy. A better one is the difference between Kepler’s laws of planetary motion and Newton’s laws of motion and gravity. Kepler’s laws described how the planets move; Newton’s laws explained why they move as Kepler described.
BA: The phenomena observed in the operant lab (e.g., shaping, reinforcement, extinction) are well-established, reproducible findings that have been applied successfully in many practical situations (more humane animal training being but one example). If PCT has any validity, it must be possible to understand these phenomena from within the PCT framework. Moreover, showing how traditional interpretations in terms of such entities as reinforcers can be reinterpreted within that framework would provide a bridge between the traditional paradigm and the PCT approach.
BA: My contribution to this discussion began when you suggested that Brembs might be “going off the rails� because of his statement that “spontaneous behavior� was used by flies to exert control. Brembs is coming from an operant conditioning point of view.  I don’t see his description as “going off the [PCT] rails,� because I see spontaneous behavior as evidence for the reorganizing process at work, selecting different behaviors until one is found that reduces whatever error is present in some perception that the fly needs to gain (or regain) control over. From Brembs’ perspective, it might look like shaping at work (via differential reinforcement), but as I’ve tried to show, there is a simple way to relate operant shaping to the process of reorganization. And the result of that reorganization process (if successful) is control.
Bruce
 I see spontaneous behavior as evidence for the reorganizing process at work, selecting different behaviors until one is found that reduces whatever error is present in some perception that the fly needs to gain (or regain) control over.Â
Where can I find an explicit model (differential equations and/or computer simulation, not a verbal description nor an interpretation of what animals do) of a simple system that, upon losing control, would by itself start to re-organize (change: reference point, the gain, etc) until it regains control and stops re-organization? In other words, explicit components that would know when control is lost, send it “up”, and then come “down” with a way to vary some para-meters, so as to try to gain control again and then stop. Thanks, AlexÂ
···
On Thu, Jun 23, 2016 at 1:55 AM, Bruce Abbott bbabbott@frontier.com wrote:
[From Bruce Abbott (2016.06.22.1955 EST)]
Â
Rick Marken (2016.06.21.1950) –
Â
Bruce Abbott (2016.20.1900 EDT)
Â
RM: That’s the mistake Skinner made. He assumed that the environment (nature) shapes behavior just like he did. But, unlike Skinner, the environment is not a control system; it doesn’t shape behavior because it has no goals (references) for how an organism should behave so it doesn’t act (as Skinner does) by “reinforcing” successive approximations to the desired (goal) behavior. This is all explained in Chapter 6 of “Controlling People”; get your copy today!!
Â
BA: Skinner did not make that mistake. Shaping as a consequence of environmental contingencies does not involve goals on the part of the environment, and Skinner did not assume that it does (see below).
Â
RM: I didn’t say that Skinner’s mistake was to think that environmental shaping involved goals like it does in his own shaping; his mistake was to think that shaping doesn’t involve goals at all whether the shaping is done by the environment or by a person (like himself).
Â
RM: Of course he didn’t explicitly assume that [shaping] involves goals on the part of the environment. He didn’t even understand that it required goals on his part either.
Â
BA: I agree with the first sentence above, and would add to it, “nor did he explicitly assume them.� I’m quite sure Skinner understood that the shaping procedure required goals on his part, or for that matter on the part of anyone engaging in the process.
Â
RM:  Skinner may have “known” that he was working to achieve goals when he was shaping behavior – that is, he may have known that shaping seemed to be purposeful – but he did everything he could to prove that purpose was actually an illusion. Indeed, he argued that his own behavior  – including the behavior of shaping behavior – was itself shaped by the environment. According to Skinner his own shaping behavior was shaped by the behavior of the animal he was shaping; he called it “reciprocal control”, remember.Â
Â
BA: Ah HA! Something we agree on! from Skinner’s point of view, shaping is analogous to evolution by natural selection, which results in organisms becoming better adapted to their environment even though purpose plays no part in the process. Indeed, Skinner would give no credit to himself for his accomplishments, contending instead that they were simply products of the environmental contingencies that had acted upon him over his lifetime. But this ignores the fact that behavior is organized in the service of control, and that what and how we control are products of both our innate natures and the environment within which our control systems must act.
Â
 BA: It’s obvious – yoou have to know what you are trying to accomplish.Â
Â
RM: It’s obvious to you but it was not obvious to Skinner or any of his followers.Â
Â
BA: As for not understanding the nature of what he was doing (control), he understood it well enough to develop the prototype of the world’s first “smart bombâ€? – one that would apply corrective actions too its fins in order to steer the bomb toward its target.Â
Â
RM: He didn’t need to understand that his own behavior was a process of control in order to train a pigeon to  keep a cursor on target by pecking. Indeed, thousands of people – psychologists, animal trainers and control engineers – were able to produce systems that control (like the pigeon in the missile controlling the cursor) without knowing that they did it by controlling those systems. Indeed, Bill Powers was the first one to understand that the behavior of living organisms IS control. So Skinner was not alone in not knowing that he was controlling. And there are still only a few people in the world who know that behavior is control. But Tim and I are trying to up that number with our book “Controlling People”. You should get a copy, or two;-)Â
Â
BA: People have been creating useful control systems for millennia without having a detailed understanding of control theory. Skinner’s “smart bomb� was to be used on enemy ships. An image of the target ship appeared on a screen and the mechanism Skinner designed changed the bomb’s fin angles so as to bring the image of the ship onto the center of the screen. There was a pair of simple feedback loops (for the horizontal and vertical fins) designed to reduce the error between the position of the image and the desired position. Skinner certainly understood that.
Â
BA: The key element that Skinner introduced was to have the location of the target on the screen indicated electronically by where on the screen a pigeon pecked. Skinner trained his pigeons to peck at the target image. The difference between this location and screen center determined the fin angles, which steered the bomb so as to zero out this difference. The pigeons were even able to distinguish the aerial image of the intended target ship from that of other ships, or in other words, to discriminate target from similar non-targets.
Â
BA: If we equate reinforcers with inputs that reduce error in a controlled perception and punishers with inputs that increase such error, then we can understand natural shaping in terms of reorganization.
Â
RM: The is no such thing as an input that reduces error? It’s the actions that affect the state of an input that reduce error (if the feedback in the loop is negative). When you say a reinforcer is an input that reduces error you are verbally turning inputs into variables that cause what the system does – the input-output model of behavior.
Â
BA: I have a hunger-control system whose reference is zero hunger. By definition, if I am hungry then there is error present in that system. If you input food to this system, hunger diminishes. Is that not “an input that reduces error?� Of course it is!
Â
BA: The food may have been consumed because the error causes food to be eaten (the action). But it is not eating (the action) that reduces the error, it is the food. Also, the food will reduce the error even if directly injected into the stomach. In that case it serves as a disturbance to the hunger control system – a helpful one in this case since it will reduce the error. So the food can reduce error (if error is present) whether it enters the control system via the system’s own actions or via an independent agent.
Â
RM: But in a control system, the inputs are controlled by the system; they don’t cause (or control or select) the behavior of the system In fact, reinforcers, like food pellets, are not inputs at all but independent events that are disturbances to controlled variables. Organisms act to prevent these disturbances from having an effect on the controlled variable. So pigeons peck keys to control their perception of getting fed by preventing the food pellets from disturbing this variable by not coming often enough. The organism is controlling its input (the state of the controlled variable); the input is not controlling (“shaping”) the organism.
Â
BA: To analyze the situation correctly we have to distinguish carefully between two situations. One is the operation of an established control system. The other is reorganization. Reorganization can create new control systems, but most of the time it acts to modify existing systems. In this discussion I have been focusing on the process of shaping and have proposed that shaping involves the reorganization of the output mechanism. If the organism is experiencing error in a controlled perception, it will perform the actions dictated by its current organization, actions that in the past have succeeded in reducing the error to acceptable levels. If these now consistently fail, other actions will be tried as reorganization varies the output function. Those actions that succeed in restoring control will be retained. That is selection by consequences.
Â
BA: In the Skinnerian system, actions are selected when their consequence is the production of reinforcers, or the prevention/removal of punishers. In PCT, actions are selected at random and retained to the extent they succeed in reducing error in a given control system. Either way, it’s selection by consequences.Â
Â
BA:  You want us to believe that by the phrase, “selecting by consequences,â€? Skinner meant that the consequences are some kind of agent actively selecting behavior. He doesn’t.Â
Â
RM: I know. And this process is not the same as the “shaping” carried out by Skinner, which involves selection of consequences (in the case of shaping, selection of a consequence of giving and withholding reinforcers, like a pigeon playing ping pong).Â
Â
BA: Wrong. It’s the same shaping process whether carried out by an external agent or as a consequence of natural environmental contingencies.
Â
BA: In natural shaping, what particular acts will produce a given reinforcer, and how efficiently they will do so, depends on the environment – what wee in PCT call the feedback function. If the organism is “spontaneously active,â€? those variant activities that chance to produce the reinforcer will increase in probability relative to other activities, and those variants of that effective act that do so more effectively will be selected over those that are less efficient.Â
Â
 RM: Was it natural shaping or shaping done by some amazing shaper that reinforced Beethoven’s “acts” (the Sonatas, quartets, Symphones, etc),  Shakespeare’s plays, Newton’s Principia, Joyce’s Ulysses?Â
Â
BA: Probably both. Given that selection by consequences is how reorganization works, one will be as hard pressed to explain these extraordinary performances from the PCT perspective as from the Skinnerian.
Â
BA: This isn’t some theoretical notion – you can obserrve the process in action.
Â
RM: Yes ,you can see it. It’s called “the behavioral illusion”; the illusion that inputs select outputs (See Powers’ reply to operant conditioners Baum and Reese starting on p. 83 of LCS I. I imagine you have already read it and dismissed it. But I think Powers made some great points, Â especially when he said: “If control-system theory does indeed describe correctly the relationship between organisms and their environments, behaviorism has been in the grip of a powerful illusion since its conceptual foundations were laid”.
Â
BA: Shaping has nothing to do with the behavioral illusion, i.e., mistaking the inverse of the feedback function (a property of the environment) for a property of the organism’s behavioral system. Shaping (natural or externally directed) determines which among various candidate output functions will be retained by a given control system.
Â
BA: What PCT offers is an explanation for the effectiveness of these consequences in bringing about changes in behavior. The language of operant conditioning describes observable, repeatable relationships, but it only describes them. PCT describes the mechanism that enforces those observed relationships.
Â
RM: Yes, it sure does. It shows that what appears to be going on – selection  BY consequences – is not what’s going on. What’s going on is selection OF consequences: control. It looks like consequences select actions but it’s actually the other way around. And the consequences that are selected are not reinforcers; they are functions of reinforcers called controlled variables.
Â
BA: So, according to you (your words), in “selection OF consequences,� actions select consequences. And those consequences are controlled variables. I don’t recall any mechanisms in PCT by which an output of a control system selects controlled variables. Please explain!
Â
BA: In the past, the differences between the Skinnerian approach and PCT have been compared on this list to the rivalry between the old Ptolemaic, earth-centered view of the heavens and the Copernican sun-centered view, but this is a poor analogy. A better one is the difference between Kepler’s laws of planetary motion and Newton’s laws of motion and gravity. Kepler’s laws described how the planets move; Newton’s laws explained why they move as Kepler described.Â
BA: The phenomena observed in the operant lab (e.g., shaping, reinforcement, extinction) are well-established, reproducible findings that have been applied successfully in many practical situations (more humane animal training being but one example). If PCT has any validity, it must be possible to understand these phenomena from within the PCT framework. Moreover, showing how traditional interpretations in terms of such entities as reinforcers can be reinterpreted within that framework would provide a bridge between the traditional paradigm and the PCT approach.
Â
BA: My contribution to this discussion began when you suggested that Brembs might be “going off the rails� because of his statement that “spontaneous behavior� was used by flies to exert control. Brembs is coming from an operant conditioning point of view. I don’t see his description as “going off the [PCT] rails,� because I see spontaneous behavior as evidence for the reorganizing process at work, selecting different behaviors until one is found that reduces whatever error is present in some perception that the fly needs to gain (or regain) control over. From Brembs’ perspective, it might look like shaping at work (via differential reinforcement), but as I’ve tried to show, there is a simple way to relate operant shaping to the process of reorganization. And the result of that reorganization process (if successful) is control.
Â
Bruce
Â
[From Bruce Abbott (2016.06.23.0905 EDT)]
···
From: Alex Gomez-Marin [mailto:agomezmarin@gmail.com]
Sent: Thursday, June 23, 2016 3:50 AM
To: csgnet@lists.illinois.edu
Subject: Re: operant behavior in model systems
I see spontaneous behavior as evidence for the reorganizing process at work, selecting different behaviors until one is found that reduces whatever error is present in some perception that the fly needs to gain (or regain) control over.
AG: Where can I find an explicit model (differential equations and/or computer simulation, not a verbal description nor an interpretation of what animals do) of a simple system that, upon losing control, would by itself start to re-organize (change: reference point, the gain, etc) until it regains control and stops re-organization? In other words, explicit components that would know when control is lost, send it “up”, and then come “down” with a way to vary some para-meters, so as to try to gain control again and then stop. Thanks, Alex
BA: Do you have Bill Powers’ book, Living Control Systems III: The Fact of Control? Chapter 8 describes a computer simulation of an arm with 14 control systems to control the angles of 14 joints. In this implementation, the output functions of all 14 control systems are connected through different randomly chosen weights to all 14 joints. The simulation manipulates the reference levels of these systems in a pattern intended to produce a well-known tai chi movement of the arm, but the initial effect is to produce a lot of flailing. However, as reorganization proceeds, the systems sort themselves out until each control system is essentially operating only a single joint, so that the tai chi movement is now recognizable. Interestingly, the system will reorganize properly whether the references are being actively manipulated or the references are fixed and smoothed random disturbances are being applied to the joint angles. In either case the control systems are actively attempting to bring the joint angles to their current reference levels while reorganization is taking place.
The simulation is not designed to lose control once good control has been established, but if it did, the reorganization process would simply do its work again.
The arm-reorganization simulation is rather limited in that reorganization only affects the weightings of the output connections to the different joints. It’s just a demonstration of principle using the simplest possible method of reorganization (hill climbing). Bill referred to this as the “e. coli method� because it appears to be the method used by the bacterium e. coli to home in on nutrient sources and avoid areas of toxicity. (A simple simulation of this e. coli behavior is also included in the package.) More sophisticated methods are of course possible. A key unresolved problem that Bill recognized is how reorganization can be limited to affect just those systems that require it without seriously disrupting the hierarchy.
With minor changes the process used in the arm-reorganization simulation could be applied to the problem of finding an effective means of controlling some perception. An animal might have several pre-organized behavioral means available to it but does not know which if any of them will be effective in establishing control over that perception. As error grows, reorganization would begin to vary the connections from the output of this system to the references of lower-level control systems that orchestrate these different behavioral acts. The external observer would see spontaneous activity of various sorts until an action was discovered that was effective in establishing control over some variable, such as the delivery of food.
Whether you have Bill’s book or not, you can download all the associated computer simulations (for Windows) at http://www.billpct.org/ The download includes the executables and Delphi source code.
Bruce
[From Rick Marken (2016.06.23.1440)]
···
Bruce Abbott (2016.06.22.1955 EST)
RM: Â … According to Skinner his own shaping behavior was shaped by the behavior of the animal he was shaping; he called it “reciprocal control”, remember.Â
Â
BA: Ah HA! Something we agree on! from Skinner’s point of view, shaping is analogous to evolution by natural selection, which results in organisms becoming better adapted to their environment even though purpose plays no part in the process. Indeed, Skinner would give no credit to himself for his accomplishments, contending instead that they were simply products of the environmental contingencies that had acted upon him over his lifetime. But this ignores the fact that behavior is organized in the service of control, and that what and how we control are products of both our innate natures and the environment within which our control systems must act.
RM: So if you know this why in the world do you keep arguing that there is something of value in the work of behaviorists like Skinner. They didn’t know they were looking at control so the observations they made were just like those made by one of the blind men in my “Blind Men and the Elephant” paper (in “More Mind Readings”); they give a completely misleading picture of the “elephant” of control – because they didn’t know that the objects of their study – the organisms in their experiments – were controllers just like they are. The whole point of PCT is that if you don’t know what control is and you don’t know that behavior is control then your “discoveries” about behavior are useless at best and misleading at worst.
RM: The is no such thing as an input that reduces error…
Â
BA: I have a hunger-control system whose reference is zero hunger. By definition, if I am hungry then there is error present in that system. If you input food to this system, hunger diminishes. Is that not “an input that reduces error?� Of course it is!
RM: Actually, it’s not. It’s a disturbance variable that, when acted on by the outputs of the hunger control system – Â digestive enzyme secretions – causes the controlled input variable – blood sugar level – to move toward the reference. And it’s not correct to say “inputs reduce error”. Error is the difference between input and reference. So error is reduced by effects on the input that move it toward the reference. Both system outputs and disturbances have effects on system inputs so either one (or both) can, at any instant, be responsible for reducing error. But it’s only the effects on input that are responsible for reducing error; not the input itself.Â
BA: The food may have been consumed because the error causes food to be eaten (the action). But it is not eating (the action) that reduces the error, it is the food.Â
RM: No, it’s the effect of system output (digestive enzyme secretion) on the food that reduces error by moving the controlled variable (blood sugar) closer to the reference. I’m sure you understand this because you have programmed all those control systems for Bill’s demos. I think you are getting it wrong here only because we are discussing it in the context of operant conditioning, which has been an important part of your career and you seem to still be controlling for it having some value. Â
Â
BA: To analyze the situation correctly we have to distinguish carefully between two situations. One is the operation of an established control system. The other is reorganization.Â
RM: Very important distinction, and I think it’s one of the main reasons that operant conditioning research is so useless (and misleading). Operant conditioners don’t make this distinction because they don’t know what behavior is. So they don’t know that what is shaped is not “output” (as you say below) but a control process. Pecking a key, for example, Â is not an output; it is a controlled consequence of output.Â
Â
BA: Reorganization can create new control systems, but most of the time it acts to modify existing systems. In this discussion I have been focusing on the process of shaping and have proposed that shaping involves the reorganization of the output mechanism. If the organism is experiencing error in a controlled perception, it will perform the actions dictated by its current organization, actions that in the past have succeeded in reducing the error to acceptable levels. If these now consistently fail, other actions will be tried as reorganization varies the output function. Those actions that succeed in restoring control will be retained. That is selection by consequences.
RM: Reorganization of the output “mechanism” is changing the output function – nature of the relationship between variations in error and actoin – not selecting new actions. This may not be the same as creating an entirely new control system but it is certainly making an existing control system control in a new way.Â
Â
RM: I know. And this process is not the same as the “shaping” carried out by Skinner, which involves selection of consequences (in the case of shaping, selection of a consequence of giving and withholding reinforcers, like a pigeon playing ping pong).Â
Â
BA: Wrong. It’s the same shaping process whether carried out by an external agent or as a consequence of natural environmental contingencies.
RM: The difference is that Skinner gives and withholds  reinforcers in order to get his perception of the behavior to a reference state; the shaping done by Skinner is purposeful.The environment gives and withholds reinforcers randomly; the shaping done by the environment is not purposeful. The shaping done by Skinner is like that done by a sculptor shaping a sculpture. The shaping done by the environment is like that done by the terrain that shapes the course of a river. Although these two kinds of shaping may seem superficially the same there’s a way to tell the difference between them. It’s done using the Test for the Controlled Variable.
Â
BA: Shaping has nothing to do with the behavioral illusion, i.e., mistaking the inverse of the feedback function (a property of the environment)Â
RM: It is one version of the behavioral illusion; the illusion that external events select the behaviors that produce them. See my “Blind Men and the Elephant” paper in “More Mind Readings”. Each of the blind men represents a different version of the behavioral illusion: the S-R illusion, the selection by consequences illusion and the computed output illusion. Each of these illusions is the basis of a major school of psychology: behaviorism, Skinnerism and cognitive science, respectively.
Â
BA: So, according to you (your words), in “selection OF consequences,� actions select consequences. And those consequences are controlled variables. I don’t recall any mechanisms in PCT by which an output of a control system selects controlled variables. Please explain!
RM: I use the phrase  “selection of consequences” to contrast the PCT view with the operant behavior view. In PCT, consequences are states of perceptions that are selected by the setting of reference signals in the organism; actions bring those perceptions to the selected state while protecting them from disturbances. In the operant behavior view, consequences select (strengthen) the actions that produce them so that the consequences continue to be produced.Â
BA: My contribution to this discussion began when you suggested that Brembs might be “going off the rails� because of his statement that “spontaneous behavior� was used by flies to exert control. Brembs is coming from an operant conditioning point of view. I don’t see his description as “going off the [PCT] rails,� because I see spontaneous behavior as evidence for the reorganizing process at work, selecting different behaviors until one is found that reduces whatever error is present in some perception that the fly needs to gain (or regain) control over. From Brembs’ perspective, it might look like shaping at work (via differential reinforcement), but as I’ve tried to show, there is a simple way to relate operant shaping to the process of reorganization. And the result of that reorganization process (if successful) is control.
RM: I agree. You can explain the phenomenon of shaping (from both the shaper and the shapee’s perspective) using PCT. How about building a model of it yourself.Â
BestÂ
Rick
–
Richard S. MarkenÂ
Author, with Timothy A. Carey, of  Controlling People: The Paradoxical Nature of Being Human.Â
RM: “The whole point of PCT is that if you don’t know what control is and you don’t know that behavior is control then your “discoveries” about behavior are useless at best and misleading at worst.”
AGM: If the whole point of PCT is to call anything that is non-PCT useless at best and misleading at worst, then our own good-willed efforts will also be useless at best and misleading at worst.…
Have a great sad #brexit day…
···
On Thu, Jun 23, 2016 at 10:43 PM, Richard Marken rsmarken@gmail.com wrote:
[From Rick Marken (2016.06.23.1440)]
Bruce Abbott (2016.06.22.1955 EST)
RM: Â … According to Skinner his own shaping behavior was shaped by the behavior of the animal he was shaping; he called it “reciprocal control”, remember.Â
Â
BA: Ah HA! Something we agree on! from Skinner’s point of view, shaping is analogous to evolution by natural selection, which results in organisms becoming better adapted to their environment even though purpose plays no part in the process. Indeed, Skinner would give no credit to himself for his accomplishments, contending instead that they were simply products of the environmental contingencies that had acted upon him over his lifetime. But this ignores the fact that behavior is organized in the service of control, and that what and how we control are products of both our innate natures and the environment within which our control systems must act.
RM: So if you know this why in the world do you keep arguing that there is something of value in the work of behaviorists like Skinner. They didn’t know they were looking at control so the observations they made were just like those made by one of the blind men in my “Blind Men and the Elephant” paper (in “More Mind Readings”); they give a completely misleading picture of the “elephant” of control – because they didn’t know that the objects of their study – the organisms in their experiments – were controllers just like they are. The whole point of PCT is that if you don’t know what control is and you don’t know that behavior is control then your “discoveries” about behavior are useless at best and misleading at worst.
RM: The is no such thing as an input that reduces error…
Â
BA: I have a hunger-control system whose reference is zero hunger. By definition, if I am hungry then there is error present in that system. If you input food to this system, hunger diminishes. Is that not “an input that reduces error?� Of course it is!
RM: Actually, it’s not. It’s a disturbance variable that, when acted on by the outputs of the hunger control system – Â digestive enzyme secretions – causes the controlled input variable – blood sugar level – to move toward the reference. And it’s not correct to say “inputs reduce error”. Error is the difference between input and reference. So error is reduced by effects on the input that move it toward the reference. Both system outputs and disturbances have effects on system inputs so either one (or both) can, at any instant, be responsible for reducing error. But it’s only the effects on input that are responsible for reducing error; not the input itself.Â
BA: The food may have been consumed because the error causes food to be eaten (the action). But it is not eating (the action) that reduces the error, it is the food.Â
RM: No, it’s the effect of system output (digestive enzyme secretion) on the food that reduces error by moving the controlled variable (blood sugar) closer to the reference. I’m sure you understand this because you have programmed all those control systems for Bill’s demos. I think you are getting it wrong here only because we are discussing it in the context of operant conditioning, which has been an important part of your career and you seem to still be controlling for it having some value. Â
Â
BA: To analyze the situation correctly we have to distinguish carefully between two situations. One is the operation of an established control system. The other is reorganization.Â
RM: Very important distinction, and I think it’s one of the main reasons that operant conditioning research is so useless (and misleading). Operant conditioners don’t make this distinction because they don’t know what behavior is. So they don’t know that what is shaped is not “output” (as you say below) but a control process. Pecking a key, for example, Â is not an output; it is a controlled consequence of output.Â
Â
BA: Reorganization can create new control systems, but most of the time it acts to modify existing systems. In this discussion I have been focusing on the process of shaping and have proposed that shaping involves the reorganization of the output mechanism. If the organism is experiencing error in a controlled perception, it will perform the actions dictated by its current organization, actions that in the past have succeeded in reducing the error to acceptable levels. If these now consistently fail, other actions will be tried as reorganization varies the output function. Those actions that succeed in restoring control will be retained. That is selection by consequences.
RM: Reorganization of the output “mechanism” is changing the output function – nature of the relationship between variations in error and actoin – not selecting new actions. This may not be the same as creating an entirely new control system but it is certainly making an existing control system control in a new way.Â
Â
RM: I know. And this process is not the same as the “shaping” carried out by Skinner, which involves selection of consequences (in the case of shaping, selection of a consequence of giving and withholding reinforcers, like a pigeon playing ping pong).Â
Â
BA: Wrong. It’s the same shaping process whether carried out by an external agent or as a consequence of natural environmental contingencies.
RM: The difference is that Skinner gives and withholds  reinforcers in order to get his perception of the behavior to a reference state; the shaping done by Skinner is purposeful.The environment gives and withholds reinforcers randomly; the shaping done by the environment is not purposeful. The shaping done by Skinner is like that done by a sculptor shaping a sculpture. The shaping done by the environment is like that done by the terrain that shapes the course of a river. Although these two kinds of shaping may seem superficially the same there’s a way to tell the difference between them. It’s done using the Test for the Controlled Variable.
Â
BA: Shaping has nothing to do with the behavioral illusion, i.e., mistaking the inverse of the feedback function (a property of the environment)Â
RM: It is one version of the behavioral illusion; the illusion that external events select the behaviors that produce them. See my “Blind Men and the Elephant” paper in “More Mind Readings”. Each of the blind men represents a different version of the behavioral illusion: the S-R illusion, the selection by consequences illusion and the computed output illusion. Each of these illusions is the basis of a major school of psychology: behaviorism, Skinnerism and cognitive science, respectively.
Â
BA: So, according to you (your words), in “selection OF consequences,� actions select consequences. And those consequences are controlled variables. I don’t recall any mechanisms in PCT by which an output of a control system selects controlled variables. Please explain!
RM: I use the phrase  “selection of consequences” to contrast the PCT view with the operant behavior view. In PCT, consequences are states of perceptions that are selected by the setting of reference signals in the organism; actions bring those perceptions to the selected state while protecting them from disturbances. In the operant behavior view, consequences select (strengthen) the actions that produce them so that the consequences continue to be produced.Â
BA: My contribution to this discussion began when you suggested that Brembs might be “going off the rails� because of his statement that “spontaneous behavior� was used by flies to exert control. Brembs is coming from an operant conditioning point of view. I don’t see his description as “going off the [PCT] rails,� because I see spontaneous behavior as evidence for the reorganizing process at work, selecting different behaviors until one is found that reduces whatever error is present in some perception that the fly needs to gain (or regain) control over. From Brembs’ perspective, it might look like shaping at work (via differential reinforcement), but as I’ve tried to show, there is a simple way to relate operant shaping to the process of reorganization. And the result of that reorganization process (if successful) is control.
RM: I agree. You can explain the phenomenon of shaping (from both the shaper and the shapee’s perspective) using PCT. How about building a model of it yourself.Â
BestÂ
Rick
–
Richard S. MarkenÂ
Author, with Timothy A. Carey, of  Controlling People: The Paradoxical Nature of Being Human.Â
Rick you are THE World Champion in misleading the whole CSGnet. After all our conversations you still don’t understand PCT.
Behavior is not control. It affects environment. See Bill’s diagram in LCS III. There is no »controlled variable« in environment so that behavior could be control. Control is not perceived from the environment, control is done in organism. I don’t know how many times you changed your mind, But if you stay on the position that »behavior is control«, then this is the worst scenario for PCT.
Best,
Boris
···
From: Alex Gomez-Marin [mailto:agomezmarin@gmail.com]
Sent: Friday, June 24, 2016 11:13 AM
To: csgnet@lists.illinois.edu
Subject: Re: operant behavior in model systems
RM: “The whole point of PCT is that if you don’t know what control is and you don’t know that behavior is control then your “discoveries” about behavior are useless at best and misleading at worst.”
AGM: If the whole point of PCT is to call anything that is non-PCT useless at best and misleading at worst, then our own good-willed efforts will also be useless at best and misleading at worst.…
Have a great sad #brexit day…
On Thu, Jun 23, 2016 at 10:43 PM, Richard Marken rsmarken@gmail.com wrote:
[From Rick Marken (2016.06.23.1440)]
Bruce Abbott (2016.06.22.1955 EST)
RM: … According to Skinner his own shaping behavior was shaped by the behavior of the animal he was shaping; he called it “reciprocal control”, remember.
BA: Ah HA! Something we agree on! From Skinner’s point of view, shaping is analogous to evolution by natural selection, which results in organisms becoming better adapted to their environment even though purpose plays no part in the process. Indeed, Skinner would give no credit to himself for his accomplishments, contending instead that they were simply products of the environmental contingencies that had acted upon him over his lifetime. But this ignores the fact that behavior is organized in the service of control, and that what and how we control are products of both our innate natures and the environment within which our control systems must act.
RM: So if you know this why in the world do you keep arguing that there is something of value in the work of behaviorists like Skinner. They didn’t know they were looking at control so the observations they made were just like those made by one of the blind men in my “Blind Men and the Elephant” paper (in “More Mind Readings”); they give a completely misleading picture of the “elephant” of control – because they didn’t know that the objects of their study – the organisms in their experiments – were controllers just like they are. The whole point of PCT is that if you don’t know what control is and you don’t know that behavior is control then your “discoveries” about behavior are useless at best and misleading at worst.
RM: The is no such thing as an input that reduces error…
BA: I have a hunger-control system whose reference is zero hunger. By definition, if I am hungry then there is error present in that system. If you input food to this system, hunger diminishes. Is that not “an input that reduces error?� Of course it is!
RM: Actually, it’s not. It’s a disturbance variable that, when acted on by the outputs of the hunger control system – digestive enzyme secretions – causes the controlled input variable – blood sugar level – to move toward the reference. And it’s not correct to say “inputs reduce error”. Error is the difference between input and reference. So error is reduced by effects on the input that move it toward the reference. Both system outputs and disturbances have effects on system inputs so either one (or both) can, at any instant, be responsible for reducing error. But it’s only the effects on input that are responsible for reducing error; not the input itself.
BA: The food may have been consumed because the error causes food to be eaten (the action). But it is not eating (the action) that reduces the error, it is the food.
RM: No, it’s the effect of system output (digestive enzyme secretion) on the food that reduces error by moving the controlled variable (blood sugar) closer to the reference. I’m sure you understand this because you have programmed all those control systems for Bill’s demos. I think you are getting it wrong here only because we are discussing it in the context of operant conditioning, which has been an important part of your career and you seem to still be controlling for it having some value.
BA: To analyze the situation correctly we have to distinguish carefully between two situations. One is the operation of an established control system. The other is reorganization.
RM: Very important distinction, and I think it’s one of the main reasons that operant conditioning research is so useless (and misleading). Operant conditioners don’t make this distinction because they don’t know what behavior is. So they don’t know that what is shaped is not “output” (as you say below) but a control process. Pecking a key, for example, is not an output; it is a controlled consequence of output.
BA: Reorganization can create new control systems, but most of the time it acts to modify existing systems. In this discussion I have been focusing on the process of shaping and have proposed that shaping involves the reorganization of the output mechanism. If the organism is experiencing error in a controlled perception, it will perform the actions dictated by its current organization, actions that in the past have succeeded in reducing the error to acceptable levels. If these now consistently fail, other actions will be tried as reorganization varies the output function. Those actions that succeed in restoring control will be retained. That is selection by consequences.
RM: Reorganization of the output “mechanism” is changing the output function – nature of the relationship between variations in error and actoin – not selecting new actions. This may not be the same as creating an entirely new control system but it is certainly making an existing control system control in a new way.
RM: I know. And this process is not the same as the “shaping” carried out by Skinner, which involves selection of consequences (in the case of shaping, selection of a consequence of giving and withholding reinforcers, like a pigeon playing ping pong).
BA: Wrong. It’s the same shaping process whether carried out by an external agent or as a consequence of natural environmental contingencies.
RM: The difference is that Skinner gives and withholds reinforcers in order to get his perception of the behavior to a reference state; the shaping done by Skinner is purposeful.The environment gives and withholds reinforcers randomly; the shaping done by the environment is not purposeful. The shaping done by Skinner is like that done by a sculptor shaping a sculpture. The shaping done by the environment is like that done by the terrain that shapes the course of a river. Although these two kinds of shaping may seem superficially the same there’s a way to tell the difference between them. It’s done using the Test for the Controlled Variable.
BA: Shaping has nothing to do with the behavioral illusion, i.e., mistaking the inverse of the feedback function (a property of the environment)
RM: It is one version of the behavioral illusion; the illusion that external events select the behaviors that produce them. See my “Blind Men and the Elephant” paper in “More Mind Readings”. Each of the blind men represents a different version of the behavioral illusion: the S-R illusion, the selection by consequences illusion and the computed output illusion. Each of these illusions is the basis of a major school of psychology: behaviorism, Skinnerism and cognitive science, respectively.
BA: So, according to you (your words), in “selection OF consequences,� actions select consequences. And those consequences are controlled variables. I don’t recall any mechanisms in PCT by which an output of a control system selects controlled variables. Please explain!
RM: I use the phrase “selection of consequences” to contrast the PCT view with the operant behavior view. In PCT, consequences are states of perceptions that are selected by the setting of reference signals in the organism; actions bring those perceptions to the selected state while protecting them from disturbances. In the operant behavior view, consequences select (strengthen) the actions that produce them so that the consequences continue to be produced.
BA: My contribution to this discussion began when you suggested that Brembs might be “going off the rails� because of his statement that “spontaneous behavior� was used by flies to exert control. Brembs is coming from an operant conditioning point of view. I don’t see his description as “going off the [PCT] rails,� because I see spontaneous behavior as evidence for the reorganizing process at work, selecting different behaviors until one is found that reduces whatever error is present in some perception that the fly needs to gain (or regain) control over. From Brembs’ perspective, it might look like shaping at work (via differential reinforcement), but as I’ve tried to show, there is a simple way to relate operant shaping to the process of reorganization. And the result of that reorganization process (if successful) is control.
RM: I agree. You can explain the phenomenon of shaping (from both the shaper and the shapee’s perspective) using PCT. How about building a model of it yourself.
Best
Rick
–
Richard S. Marken
Author, with Timothy A. Carey, of Controlling People: The Paradoxical Nature of Being Human.
[From Bruce Abbott (2016.06.24.1145 EDT]
Rick Marken (2016.06.23.1440) –
Bruce Abbott (2016.06.22.1955 EST)
RM: … According to Skinner his own shaping behavior was shaped by the behavior of the animal he was shaping; he called it “reciprocal control”, remember.
BA: Ah HA! Something we agree on! From Skinner’s point of view, shaping is analogous to evolution by natural selection, which results in organisms becoming better adapted to their environment even though purpose plays no part in the process. Indeed, Skinner would give no credit to himself for his accomplishments, contending instead that they were simply products of the environmental contingencies that had acted upon him over his lifetime. But this ignores the fact that behavior is organized in the service of control, and that what and how we control are products of both our innate natures and the environment within which our control systems must act.
RM: So if you know this why in the world do you keep arguing that there is something of value in the work of behaviorists like Skinner. They didn’t know they were looking at control so the observations they made were just like those made by one of the blind men in my “Blind Men and the Elephant” paper (in “More Mind Readings”); they give a completely misleading picture of the “elephant” of control – because they didn’t know that the objects of their study – the organisms in their experiments – were controllers just like they are. The whole point of PCT is that if you don’t know what control is and you don’t know that behavior is control then your “discoveries” about behavior are useless at best and misleading at worst.
BA: I keep arguing that there is something of value in the work of behaviorists like Skinner because there is something of considerable value in the work of behaviorists like Skinner. There is a significant difference between behaviorist and PCT explanations for experimental findings from the operant laboratory, it is true. Behaviorist explanations avoid reference to theoretical entities that must be inferred from the data. They focus instead on discovering regularities in relationships between objectively observable environmental conditions to which an individual is exposed and the individual’s behavior. The phenomena revealed by this approach are real, reproducible, and important. They are not “useless at best and misleading at worst.� What PCT offers is not a refutation of behaviorist findings but a deeper understanding of them based on the application of control theory.
RM: The is no such thing as an input that reduces error…
BA: I have a hunger-control system whose reference is zero hunger. By definition, if I am hungry then there is error present in that system. If you input food to this system, hunger diminishes. Is that not “an input that reduces error?� Of course it is!
RM: Actually, it’s not. It’s a disturbance variable that, when acted on by the outputs of the hunger control system – digestive enzyme secretions – causes the controlled input variable – blood sugar level – to move toward the reference. And it’s not correct to say “inputs reduce error”. Error is the difference between input and reference. So error is reduced by effects on the input that move it toward the reference. Both system outputs and disturbances have effects on system inputs so either one (or both) can, at any instant, be responsible for reducing error. But it’s only the effects on input that are responsible for reducing error; not the input itself.
BA: Actually, it is. It is correct to say “inputs reduce error.� All you’ve done above is elaborate how one kind of input (food) does so. Food in the stomach reduces hunger. Stating how that happens does not refute the fact that it does.
BA: The food may have been consumed because the error causes food to be eaten (the action). But it is not eating (the action) that reduces the error, it is the food.
RM: No, it’s the effect of system output (digestive enzyme secretion) on the food that reduces error by moving the controlled variable (blood sugar) closer to the reference. I’m sure you understand this because you have programmed all those control systems for Bill’s demos. I think you are getting it wrong here only because we are discussing it in the context of operant conditioning, which has been an important part of your career and you seem to still be controlling for it having some value.
BA: Yes I do understand how food input to the digestive system reduces hunger, but again, adding details about how that happens does not refute the fact that food reduces hunger. As for my training in the experimental analysis of behavior (the name Skinner gave to his approach), it is because I am intimately familiar with that field (and now with PCT) that I feel I can speak with some authority concerning the relationship between this area of scientific research and PCT.
BA: To analyze the situation correctly we have to distinguish carefully between two situations. One is the operation of an established control system. The other is reorganization.
RM: Very important distinction, and I think it’s one of the main reasons that operant conditioning research is so useless (and misleading). Operant conditioners don’t make this distinction because they don’t know what behavior is. So they don’t know that what is shaped is not “output” (as you say below) but a control process. Pecking a key, for example, is not an output; it is a controlled consequence of output.
BA: What you believe operant conditioners know or don’t know is irrelevant to this discussion. My argument about how PCT relates to the objective observations of environment-behavior relationships obtained in the operant lab does not depend on what operant conditioners know.
BA: Reorganization can create new control systems, but most of the time it acts to modify existing systems. In this discussion I have been focusing on the process of shaping and have proposed that shaping involves the reorganization of the output mechanism. If the organism is experiencing error in a controlled perception, it will perform the actions dictated by its current organization, actions that in the past have succeeded in reducing the error to acceptable levels. If these now consistently fail, other actions will be tried as reorganization varies the output function. Those actions that succeed in restoring control will be retained. That is selection by consequences.
RM: Reorganization of the output “mechanism” is changing the output function – nature of the relationship between variations in error and actoin – not selecting new actions. This may not be the same as creating an entirely new control system but it is certainly making an existing control system control in a new way.
BA: So, if the hungry rat in an operant chamber finds that lever-pressing no longer produces food, abandons lever-pressing in favor of exploration, and ultimately discovers that pulling on a chain produces the food, you say that no new action has been selected. You really are grasping at straws here.
RM: I know. And this process is not the same as the “shaping” carried out by Skinner, which involves selection of consequences (in the case of shaping, selection of a consequence of giving and withholding reinforcers, like a pigeon playing ping pong).
BA: Wrong. It’s the same shaping process whether carried out by an external agent or as a consequence of natural environmental contingencies.
RM: The difference is that Skinner gives and withholds reinforcers in order to get his perception of the behavior to a reference state; the shaping done by Skinner is purposeful.The environment gives and withholds reinforcers randomly; the shaping done by the environment is not purposeful. The shaping done by Skinner is like that done by a sculptor shaping a sculpture. The shaping done by the environment is like that done by the terrain that shapes the course of a river. Although these two kinds of shaping may seem superficially the same there’s a way to tell the difference between them. It’s done using the Test for the Controlled Variable.
BA: The “shapingâ€? done by the animal breeder is purposeful – the breeder has some variant characteristic that she has chosen to promote through selective breeding. The “shapingâ€? done by the environment is not purposeful – the variant characteristics that are “promotedâ€? by the environment are those that simply give its possessors an edge in the race to survive and successfully reproduce. Nevertheless, selection is at work in both cases. The same is true for shaping done purposefully by an animal trainer and shaping that emerges from natural environmental contingencies. All the trainer does when shaping a behavior is manipulate the environmental contingencies. In natural shaping the contingencies are simply those found in the natural environment surrounding the individual. It’s the same underlying mechanism at work in either case that produces the changes in behavior. The test for the controlled variable would disclose what was motivating the changes in either case – what control system was experienciing the error that was driving reorganization. Identifying the controlled variable does not refute that shaping is going on.
BA: Shaping has nothing to do with the behavioral illusion, i.e., mistaking the inverse of the feedback function (a property of the environment)
RM: It is one version of the behavioral illusion; the illusion that external events select the behaviors that produce them. See my “Blind Men and the Elephant” paper in “More Mind Readings”. Each of the blind men represents a different version of the behavioral illusion: the S-R illusion, the selection by consequences illusion and the computed output illusion. Each of these illusions is the basis of a major school of psychology: behaviorism, Skinnerism and cognitive science, respectively.
BA: So I was right – the behavioral illusion hhas nothing to do with shaping. The phrase “behavioral illusionâ€? has a specific definition in PCT that those other things you mention do not fit.
BA: So, according to you (your words), in “selection OF consequences,� actions select consequences. And those consequences are controlled variables. I don’t recall any mechanisms in PCT by which an output of a control system selects controlled variables. Please explain!
RM: I use the phrase “selection of consequences” to contrast the PCT view with the operant behavior view. In PCT, consequences are states of perceptions that are selected by the setting of reference signals in the organism; actions bring those perceptions to the selected state while protecting them from disturbances. In the operant behavior view, consequences select (strengthen) the actions that produce them so that the consequences continue to be produced.
BA: Earlier in this post I noted the importance of distinguishing the operation of an established control system and reorganization. Remember, we are discussing shaping, which applies to the latter. So why do you revert to discussing “selection by consequences� in the context of the operation of an established control system?
BA: But I do want to say something about your statement. You say, “in PCT, consequences are states of perceptions that are selected by setting the reference signals in the organisms . . .� That is not the selecting we are discussing here. In discussing shaping, we’ve been discussing the selection of different means to achieve the same perceptual ends, not the selection of different levels of the perceptual signal.
BA: My contribution to this discussion began when you suggested that Brembs might be “going off the rails� because of his statement that “spontaneous behavior� was used by flies to exert control. Brembs is coming from an operant conditioning point of view. I don’t see his description as “going off the [PCT] rails,� because I see spontaneous behavior as evidence for the reorganizing process at work, selecting different behaviors until one is found that reduces whatever error is present in some perception that the fly needs to gain (or regain) control over. From Brembs’ perspective, it might look like shaping at work (via differential reinforcement), but as I’ve tried to show, there is a simple way to relate operant shaping to the process of reorganization. And the result of that reorganization process (if successful) is control.
RM: I agree. You can explain the phenomenon of shaping (from both the shaper and the shapee’s perspective) using PCT. How about building a model of it yourself.
BA: Well, I’m pleased that you agree. I do plan to build such a model. However, I have at least two other pressing projects ahead of it, and in fact I’ve already devoted too much time to this discussion given that deadlines are looming. So I think I’ll leave it at that for now.
Bruce