How behaviours stored in the brain

CSGnet Archive CSG2006
1

[From Bill Powers (2006.06.19.0730 MDT)]

Wang Bo (2006.06.18) --

What's your opinion on how behaviours stored in the brain?
A series of supposed inputs and outputs��
Or a series of intended input?

According to PCT, it seems to squints towards the second one. A sequence of intended inputs stored in brain. When the head of chain was actived(by some stimulus), every node generates output according to the intended input in it.

Or you have other ideas?

Yes, you are correct that PCT would be consistent with the second one. As you probably know, the control of sequences is just one level of organization that has been proposed. There are ten others, some at higher levels and others at lower levels. Only the lower levels have been put to any serious tests -- the rest are still only hypotheses based on informal observations.

However, the idea that an external stimulus activates a stored sequence is different from the PCT idea. In PCT, stimuli do not cause behavior, but just the opposite. The organism produces behavior (actions, movements, physical forces) as a means of causing perceptions to occur, and it adjusts the behavior to make the perceptions match internal "reference signals". Because the behavior acts on perceptions through the external world, and perceptions represent the external world, we can often observe aspects of the environment bing controlled when we see another organism behaving, expecially when we perceive in the same way that the observed organism perceives.

Independent external influences can change one or more perceptions that an organism is controlling, by acting on the world that the organism is sensing. Such influences are called disturbances. If a perception is being controlled, the organism will act to prevent it from changing, and this action will appear to be a "response" to a "stimulus". This is how the appearance of stimuli causing responses is explained in PCT. We identify such "stimuli" as disturbances of controlled variables.

At the sequence level of organization, the theory says, a sequence of perceptions would be stored as potential reference signals, like memories of the elements of the sequence. When a higher system specifies that a certain sequence is to be perceived, the organism will act (through lower-level systems) to create that sequence, matching the perceived sequence to the reference sequence. But in general, this will not result in a specific sequence of behaviors -- postures, movements, or forces -- because the environment normally is changing. Repeating the same sequence of actions will not cause the same sequence of results to occur. In order for a specific sequence of perceived results to occur, the actions must vary so as to compensate for variations in the environment. This is how the subject of negative feedback control arises in PCT. Only a negative feedback control system is able to alter its actions to keep the perceived result the same when unpredictable and invisible disturbances occur. Systems based on stimulus and response cannot do this.

I hope this is not ten times as much information as you wanted, and that I have at least partially answered your question. I would like to know what your interest in PCT is.-- can you tell us a little about yourself?

Best regards,

Bill P.

2

[From Bill Powers (2006.06.21.0545 MDT)]

Does it means like this? A sequence of "reference signals" stored on the X axis, and when it's implemented along the X axis, actions generated on the Y axis make the controlled variable approach the reference one. Meanwhile, there's no fixed plans, actions adjust the controlled variable against disturbance. According to my understanding, I drew a figure (attach file).

I'm not sure what your diagram means. Where do the parts of the sequence that have already been done exist? I think you're drawing what goes on in the mind of an observer, not what goes on in the behaving system. In the behaving system, there is never anything but present time. Past and future are in the imagination of the observer of the system.

In PCT there are levels of organization. At one level, a stored sequence may exist, and elements of that sequence are picked one at a time to serve as directives to the next level down: "System A, create this amount of your perception while System B creates another amount of its perception. Then system A create a different amount of your perception, and System C create this amount of your perception, while System B is turned off."

Nothing is said to any lower system about HOW it is to create a certain amount of the perception it controls. How it is done depends on what the current state of that perception is (if it is already in the specified state no action needs to be taken) and on the sum of all disturbances tending to change that perception. This is why you can't plan actions. You can plan only outcomes, only perceived results.

This kind of relationship applies at every level, until we get to the level where the action of the system consists of creating a muscle tension. That is the only level of system that acts on the outside world.

The following text has nothing to do with PCT, it's my brief statement. So to those who are not interested in me, I'm sorry for disturbing you.:slight_smile:

Using genetic algorithm to generate programs won't be the best choice. It costs creatures lots of time and bodies in evolution.

I agree! Are you familiar with the process I call "E. coli reorganization"? In a system with only two degrees of freedom, it is about 100 times as efficient as natural selection, while still using only random changes. It becomes even more efficient relative to natural selection when the number of degrees of freedom increases, by roughly an order of magnitude with each new DF. It would be interesting to see it implemented in a self-organizing computer program.

If we want to build a self-programming machine, a new structure is needed. So I'm thinking of how behaviors stored in our brains since they're a kind of programs. The program is a set of controlling, isn't it? I turned to PCT for help, after Archy (little bug Mr. Kennaway made) came across me.

Some programs control, but others simply execute steps blindly. Controlling requires comparing the current state of the perceived world with a reference state and acting on the basis of the difference. In computers, this comparison process is usually a logical test (is A greater than B?) but it can also be quantitative (B - A = 2.38452). These tests are what turn a sequence into a program, if we want to define a program as a network of tests.

Remember that behaviors, meaning actions or outputs, are not what control systems produce. Control systems produce results, consequences, outcomes, and they do so by varying their actions or outputs in any way required.

I'll use a kind of nodes, which is stronger than neurons and in lower level than agents, to be the basic unit of this new structure. A network links one node to another. Every node has a special variable called "temperature". To those perceptual nodes like sensors, their temperature is used to sign how much they are different from others. To those control nodes, the temperature is used to sign how much they affect on the controlled variables.

I advise against using terms like "temperature" which are not meant to be understood literally. You do not mean that a node gets hot or cold. What you're describing is a function of some kind, so why not just define it as a function? I think metaphors cause more trouble than they are worth. Since they are not meant literally, different people can be reminded of different things by them, and you end up having to correct their understanding of what you mean.

You will find that in PCT, the basic unit of organization is a negative feedback control system. That may be what you are looking for: more than a neuron, less than an agent.

So, the sequence I referred to before is not a fixed chain. After a node was implemented, it will boost the temperature of next step. In this way, the behavior may be broken in the half way since there's other disturbance on the temperature. So, the "activation" referred to in the last mail related to this thought.

I still think it would be better to substitute a specific model for metaphor. What I like is not what everybody likes, naturally.

Now I'm a freshman at Shanghai Jiao Tong University, P.R.China. So, what I said here seems pompous and naive. Please forgive me for my impertinence.

How can you have new thoughts if you worry about how you sound to other people? You must be the main critic of your own thoughts -- you're not going to take other people's criticisms as seriously as you take your own, are you? Go ahead and think up crazy ideas. You can always throw away the ones that don't work. The world knows about only a tenth of my great ideas. I am very glad for that, but I am also glad that I kept a few of them.

I repeat what Rik Marken said: I hope you can come to the CSG meeting in Guangzhou.

Best,

Bill P.

3

[From Bill Powers (2006.06.22.0712 MDT)]

Frank Wang (2006.06.21) --

��m not familiar with ��E. coli reorganization��. After looking it up,

there��s a kind of protein related to that. Did you mean the control

of flagellums on bacterias?

No, I mean the kind of reorganization that is part of my model of behavior. I am beginning to think that the problem is that you haven't read my books or any of the other material on PCT that is available. Try this link:

http://www.perceptualcontroltheory.org/

I think we will communicate better if you become familiar with PCT first.

My 1973 book, "Behavior: the control of perception" has been translated into Chinese by Prof. Zhang Hua Xia and Prof. Fan Dong Ping of South China Normal University in Guangzhou. It was published last year. I'm sure you can obtain a copy from either of these people.

Best regards,

Bill Powers