[From Bill Powers (2009.11.10.0912 MDT)]
Dick Robertson,2009.11.10.0955CST et. al. –
Since, as everyone seems to agree, the last two digits of one’s SS number
have nothing to do with judgements of worth or magnitude of something
else, if the SS does have an effect it must be of the nature of a
disturbance. We always show disturbances as acting directly on the
controlled variable, because that makes the equations the simplest, but
any influence that comes from outside the control loop and independently
alters any signal or variable in the loop will end up disturbing the
controlled variable relative to the reference level of the
moment.
The most likely place for an external influence like a social security
number to have its effect is on the perceptual side, since it enters as a
perception. In the brain, perceptual input functions are probably located
near each other in sensory nuclei, as are output functions in motor
nuclei (see one of my Byte articles comparing the conventional way we
draw the control hierarchy, and next to it the more probable arrangement
of functions in the brain. See also B:CP, the chapter on Premises, Figs.
3.11 and 3.12, showing how the many-to-one representation of input
functions that we use is simply another way to draw the more
anatomically-likely way in which different perceptual signals are derived
from the same set of inputs from lower levels).
So there are potentially several ways in which activities in one control
loop can “leak” effects into parts of other control loops that
are closely adjacent in the brain. Gavin Ritz is justified in guessing
that these “priming” effects or disturbances can occur in
several different functions of the loop.
Before everyone starts putting a lot of time and effort into experiments
with priming, however, I think it would be a good idea to ask about the
magnitude and importance of the effect and the circumstances under which
it occurs. It seems that the AReilly experiments belong to a genre of
psychological experiments in which the stated goal is the
“prediction and control of behavior toward socially accepted
norms.” In many of the examples cited, we could alter the final
phrase to “toward increasing the profits of
businesses.”
If one could predict biases in people’s judgments or preferences that
result from the local stimulus context, one could stock the shelves with
a mix of products that would leave fewer items unsold. If, in addition to
that, one could manipulate the stimulus context by some inexpensive means
such as playing different kinds of music, using different paint colors on
the walls, or wafting different odors into the air, one could steer the
customer’s preferences toward the more profitable items. In large retail
operations with low profit margins such as supermarkets, even a very
slight alteration in the distribution of purchases could make major
differences in the net income of the store. So there is a considerable
motivation to look for even minor and unimportant effects – they may be
minor and unimportant to the customer but exceedingly important to the
businessman.
To the theoretician whose motives are concerned only with an increase of
knowledge, minor effects are of little interest, since there are so many
major effects remaining to be understood. Crosstalk between components of
different control systems most probably exists, and in pathological case
may have serious effects (for example, synaesthesia). However, if the
whole hierarchy is to function effectively, irrelevant influences of one
perception on another, or of one output on another at the same level,
must be considered as flaws in the control systems. Normal reorganization
should tend to remove such crosstalk.
As we have demonstrated conclusively, it is important that
multidimensional control be carried out in such a way that variations of
perceptions in any one dimension can occur without resulting in changes
in any other dimensions. We have discussed that on CSGnet under the
heading of “orthogonality.” If there is crosstalk that results
in mutual disturbances, energy will be wasted as each control system is
forced to produce more output to cancel the effects of the other control
systems. Demo 8-1 in LCS3 showes how reorganization can restore a
multidimension collection of control systems to orthogonality, and how
that produces skillful control of an arm with 14 degrees of freedom,
starting with a total lack of orthogonality.
However, orthogonality is never perfect, so even after reorganization
there are slight effects of controlling one variable on the states of
other variables, and the disturbed systems respond with changes in their
outputs that oppose those disturbances. This might look like a
“priming” effect, and indeed might be an explanation of some
observations of that effect (though we can’t rule out memory association
as an explanation in some cases). In extreme cases, we can get a person
to see a word as colored blue just by using a number-word, such as
“five.” Numbers have nothing to do with colors, but to some
people suffering from synaesthesia, they have a very distracting effect
on color perception (which perception David Goldstein and I have shown to
be quite normal when the interfering words or letters are not
introduced).
I propose, therefore, that the phenomena of interest here are probably
present to some degree in every person, that they represent a degree of
nonorthogonality of input and/or output functions, and that in normal
people are no interest except to those who can achieve some commercial
advantage from them. From the standpoint of people who show large amounts
of these interfering effects, the main interest would be in finding ways
to help them reorganize to remove as much of the effect as
possible.
It would be interesting to know whether the AReilly phenomenon would also
be seen if reversed: could people be induced to state the last two digits
of their social security number incorrectly? That would be a great
disservice to them, so perhaps exactly this experiment should not be
done.
Best,
Bill P.