Epicycles and fundamentals

[Rick Marken (960126.1945)]

Bruce Abbott (960126.2040 EST)--

Yes. It [Killeen's model] really is a nice Ptolemaic theory of behavior,
replete with all the right cycles, epicycles, and eccentrics to fit the
observations, and just as fundamentally wrong. As I described to Rick, it
can, I believe, even give a reasonable account of the effect of "free"
incentive deliveries.

Cycles and epicycles were introduced into Ptolemaic theory _after_
observations (like the retrograde motion of Mars) were made that did
not fit the simple geocentric model. You have managed to introduce an
"epicycle" in advance of the observations.

But this is a prediction about average rates; methinks an exploitable
difference between the Killeen model and a control model may lie in
the details, details that we will be recording in our studies.

You seem to be intentionally looking for the trees (details) in order to
avoid the nice view of the forest. We don't need to go into the details of
Killeen's model to find a prediction that is "exploitable"; the prediction
that is "exploitable" is the one that would require the addition of an
"epicycle" fix if it failed.

Killeen's model makes a prediction that, in all likelihood, does fail;
the prediction about the effect of arbitrarily added incentives. The
prediction of Killeen's model, sans your added epicycle (see Note 1)
is that arbitrary addition of incentives will lead to an _increase_ in
responding. If we do this experiment and find that responding _decreases_
(as predicted by PCT) then Killeen's model will only work with the addition
of your epicycle (Note 1). The need to add this epicycle is what suggests
that Killeen's model is fundamentally wrong (what is fundamentally wrong,
of course, is the assumption that incentives cause responding; the correct
assumtion is that responding controls incentives).

The thing to do now is some observing; specifically, observing what
happens to behavior when arbitrary non-contingent incentives are
provided in a fixed ratio operant conditioning experiment. Can we start
doing some research already?


Note 1.

Bruce Abbott's proposed epicycle to be added to Killeen's model of operant
behavior in case arbitrary, non-contingent addition of incentives leads to a
decrease rather than to an increase in behavior:

If you begin to offer "free" incentives, you "strengthen" the linkage
between the incentives and the behaviors that immediately preceded
those incentives, so that the incentives begin to "incite" these other
activities rather than lever- pressing. But keypecks continue to be
followed on occasion by incentive delivery as well, so they continue to be
"incited," although to a lesser degree because much of the incited activity
is now being expended in behaviors other than keypecking.