Hi Vyv
Vyv: Thank for your reply, very helpful stuff.
Â
RM: And thanks for your reply to my reply.
···
RM: It may be just self- interest that makes me see the relevance of basic PCT science to the practice of MOL but I do believe that a good understanding of PCT science is as important to being a good MOL practitioner as a good understanding of biological science is to being a good medical practitioner. I think Tim is living evidence that that belief is not completely crazy.Â
RM: … I think the main problem with the idea that prediction is involved in control (from a PCT perspective) is that it leads to a conception of control as a process of output generation rather than control of input. This is not a problem when one is designing artificial control systems but it is a huge impediment to understanding the behavior of living control systems.
VH: It can be a problem for artificial control systems too can’t it? The control of input organisation that evolution produced is so much more efficient than control of output.
RM: It’s not a problem for artificial control systems because those systems are negative feedback systems (or they wouldn’t control) and negative feedback systems are control of input systems; that’s how they work. A thermostat controls its input, which is the temperature in the air as perceived in terms of the state of an electrical current (on or off) determined by the circumference of a bimetallic strip. The cruise control in your car controls its input, which is the speed of rotation of the wheels as perceived in terms of a "…speed signal from the rotating drive shaft." Engineers didn’t think of a negative feedback control as control of input because, from their perspective, they wanted the system to control a variable out in the environment (air temperature, wheel speed). So their diagrams always showed what we would call the controlled input as a controlled output variable. Psychologists who started to apply control theory to behavior liked this mapping of the control diagram to behavior because it fit with their S-R view of how behavior worked. So what was the “set point” or “reference” in control diagrams was taken as the stimulus (S) that caused the controlled “output” (R). Bill Powers saw that this mapping of the control diagram to behavior was wrong and misleading when applied to living systems. He was able to do this because he saw that behavior was itself a process of control. So while this mislabeling of controlled input as controlled output had no effect on the ability of engineers to design artificial control systems, as Bill says in his 1978 Psych Review paper that is my bible, it drags a red herring across the path of progress.Â
Â
Vyv: Rupert Young’s robots are a example of that I think…
RM: I’m afraid all robots that control – and there are a ton of them – are organized around the control of input. I think Rupert can make an enormous contribution to robotics development by showing how realistic, complex behavior depends on developing means of sensing variations in the kinds of high level perceptual variables (inputs) that living systems control.Â
Â
VH: There was a bomb sight developed in the US during WW2 that was supposedly much more accurate than anything before it (the Norden). It used an analogue computer to predict where bombs would land as a function of airspeed, height etc. They’d tested it mostly in ideal conditions in the southern US (nice weather, no clouds, well lit etc). When it was brought into a Northern Europe theatre the weather was so windy and conditions poor that it was really not effective at predicting where bombs would land at … didn’t stop them dropping bombs mind!!
RM: This is, indeed, an open loop system so it’s not a control system. The problem, as you note, is that the bomb’s trajectory is ballistic – there is no correction for disturbance during its fall. Once they developed bombs that could sense their relationship to the target throughout their flight they had the basis for a control system known as a “guided missile”. I think it would be pleasantly ironic if PCT will let us use control theory to help us figure out ways to minimize the conflicts that result in the use of weapons like these that were perfected using control theory.
Best
Rick
Â
RM Another problem is that prediction is computationally intensive, making it unlikely that it is involved in the controlling done by living control systems.  And still another problem (which I think you mentioned, Vyv) is that prediction is typically unfeasible (you can’t predict what you can’t sense, such as many of the disturbances that affect controlled variables) and unnecessary.Â
VH: That’s a fascinating point. I guess also that living control systems don’t work with binary circuits like in man made computers. . I find the Premises chapter in B:CP so interesting but have no ideal if the way he proposes neural computers work has any relationship with what we now know in neuroscience 45 years later… I wish I did.
RM:  Vyv suggests that it’s the perception of a program “IF there’s a duck THEN point barrel slightly in front of it’s location” which I think is basically right. I’d say it’s more like “IF you are shooting gun X THEN point barrel Y amount in front of target”. I don’t think what matters is the target (duck or clay pigeon); what matters is some familiarity with the gun. That familiarity will lead you to set the reference for the lower level perception of the the distance by which the barrel leads the target by Y amount.Â
VH: Sounds right to me…
RM: Anyway, one last point. We obviously do predict and plan. But, again as I think Vyv mentioned, this is all done in imagination before one actually starts controlling. Then you control for the barrel leading the target by Y amount, you are not controlling an imagination; you are controlling an actual perception of the distance between barrel and target. What this distance should be is likely a result of mental calculations and reckoning;
VH: I think, in a human being, would be done by trial and error rather than computation. I mean your average amateur hunter plays around with the reference for the distance between duck and barrel. If he/she consistently gets hits then the reference distance will stay the same if not it’ll vary a bit (e’coli)? This produces a control organisation that looks like prediction but isn’t?
RM: but all that is done in imagination before you actually control for the perception of the relationship between barrel and target.Â
Vyv: Definitely sounds plausible, so the shooter might try out that in the imagination after visualising the trajectories? That’d save a lot of reorganisation wouldn’t it? In Gary Cziko’s Things We Do he describes the detour task where an animal can move to get food but a barrier is placed in it’s way. Chickens tend to frantically (like Thorndikes cat) struggle at the barrier until they find a way round. Monkey’s look at it and calmly walk around. Some mammals have this means of cutting short the reorganisation in finding ways to control their perceptions.
RM: Maybe it’s fair to say that, from a PCT perspective, “prediction” involves controlling imagined perceptions; “controlling” involves control of actual perceptions.
VH: The word prediction may be broader. I might predict by following a program. If I want to know how long it’ll take me to get to a place I have to carry out a simple arithmetic distance and average speed. This isn’t pure imagination because I might need to consult a map to get the information I need, write down figures on a pad etc. In this case there is a blend of controlling imagined perceptions and actual perceptions?
VH: I guess that’s why I like BP approach, the terms in PCT are not so murky. I notice terms like ‘action’ or ‘goal’ or ‘prediction’ are not really used much. I guess because they cause problems and misunderstandings fairly easily!
All the best
Vyv
Best
Rick –
Richard S. MarkenÂ
"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
                --Antoine de Saint-Exupery
        =---------------------------------------------------------------------------
Post your message to the list by sending it to mol@mail-list.com.
To contact the list owner, send your message to
mail-list.com 1302 Waugh Dr. #438 Houston, Texas 77019 USA
To unsubscribe, switch to/from digest, get on/off vacation, or change your email address, click here.
–
Richard S. MarkenÂ
"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
                --Antoine de Saint-Exupery