Selection of consequences

[From Rick Marken (970115.0915)]

For those of you with access to Java enabled browsers, I have
completed another demo (on "Selection of consequences": the
notorious E. coli demo) and made it available at my demo site:

I have also made the Java source for this demo avaiable because
I think it provides a nice framework that can be used by others
who want to write their own Java demos or experiments.

Again, comments or suggestions would be appreciated.



[From Rick Marken (941031.1000)]

Bill Powers (941031.0700 MST) to Bruce Abbott (941030.1700 EST) - -

Wonderful post. As you note, Bruce's model is a control model, controlling
the input variable, dNut. The crucial point, which you describe so well, is

It isn't dNut that determines where the drop in tumbling rate will
occur, but the reference signal.

So dNut (a consequence of tumbling) doesn't select behavior (tumbling or
not); it is selected by behavior (at least, in the region of dNut where there
is control). In Bruce's control model (which is similar to the one Bill and I
used in our analysis of the E. coli effect with people) dNut does not act as
a reinforcement (if a reinforcement is something that strengthen's or, at
least, affects the behavior that produces it); it is a controlled perceptual

So, is this the way reinforcement works. Bruce? And, if so, why aren't EAB
researchers talking more about reinforcement as controlled perceptual input?