[From Rick Marken (941208.1510)]
Bruce Abbott (941208.1000 EST)--
In other words, because string-pulling leads to the consequence of affecting
an intrinsic variable, it gets selected and "connected," to use your term.
I corrected myself on this. I should have said that, because string pulling
ends up reducing intrinsic error, the string pulling control system does not
get reorganized away (or "unconnected from the other systems in the
hierarchy). This is the "figure ground" difference Bill was talking about.
According to Thorndike, consequences strengthen the change in the behavior
that produced them; according to PCT, consequences (effects on intrinsic
varaibles) stop futher changes in behavior if the consequences reduce
intrinsic error.
This response misses the mark. I did not ask what would happen if the cat
could no longer get sensory feedback as to whether it had pulled the string,
but rather what would happen if it could not learn of the consequence of its
behavior--the opening of the door. It pulls on the string and--nothing seems
to happen. Would string-pulling then get "built and/or connected to existing
control systems"?
Your answer has to be, I would think, "no."
The answer depends on whether string pulling ultimately has an effect on
the cat's intrinsic error. If the cat had absolutely no perception indicating
that the door is open -- no sound, no smell, no signt, etc -- the cat would,
indeed, be unlikely to get to the food, consume it and reduce intrinsic
error; but it is still POSSIBLE. The cat (as part of its random
reorganization) might build a control organization (hierarchical
arrangement of control systems) that pulls the string and the moves the cat
relative to the string in the direction of the food. Because the string has
opened the door, the cat gets to the food.
Congratulations! You have just described the law of effect.
If the law of effect says that "the organism can only learn behaviors that
have the effect of reducing intrinsic error" then, indeed, I have. So, if we
know that food (X) (which reduces intrinsic error) is caused by behaviors A,
B and C but not by behaviors E, F and G, then my "law of effect" says only
that an organism could learn to get X using behaviors A,B or C but it can't
learn to get X using behaviors E, F or G. This law of effect doesn't say that
the effect (X) causes the behavior that produces it; it merely says that a
behavior must have the effect of producing a particular consequence (X) if it
is to be learned as a means of controlling that consequnce.
The flaw in the real "law of effect" is in assuming that a causal path exists
even though it is not observed; we can see that A, B or C cause X but
behaviorsists IMAGINE that X causes the next A, B or C.
Best
Rick