[From Bill Powers (980426.1105 MDT)]
Fred Nickols (9804236)
I guess I was so taken aback by the equating of "expectation" and "reference
level" that I posed my question too hastily. (I was taken aback because to
date I've understood that expectations do not equal reference levels.)
How would you define an expectation in the HPCT framework? Or under any
other model?
In any event, I'll try again.
Would the making of a perception of the world we experience into a reference
level for it qualify as a partial definition of learning?
I'm still not sure how to answer your question. Are we creating a new
meaning for the word "learning" here, or are you asking if making a
perception into a reference level corresponds to some already-defined
phenomenon that we already mean by the word learning? Learning isn't really
a technical term as it's already used -- when I use it I mean to refer
loosely to changes in habitual ways of behaving, without saying whether the
changes amount to reorganization, acquisition of new information, or some
higher-order system acting in an already-learned way to alter how a
lower-order system reacts to disturbances.
If you see a traffic light showing yellow, this leads to an expectation
that the red light will be next in the sequence, followed by releasing the
brake and stepping on the accelerator pedal. Is that "learning?" I wouldn't
think so. But does it involve setting up a reference signal? I think it
does; the next color that _should_ appear is red. If we see green instead,
we would be very surprised, and would see this as an error in the
"expected" sequence. Traffic lights do not go green, yellow, green. The
second green would be a sequence error, showing that we have a reference
condition set up defining the "right" sequence.
But that's not quite what you're asking, is it? When we initially form this
expectation, we don't know what the "natural" sequence of colors of a
traffic light is. We have to learn the sequence by observing it. So we
start out with a perception of a sequence that gradually becomes familiar,
meaning we remember its having occurred before. When we have seen this
sequence often enough, we come to "expect" it. So I guess the answer to
your question is "yes," because we form this expectation only after having
experienced the sequence enough times. Once it exists in memory, a
higher-level system can then try to reproduce the same experience. It does
this, under PCT, by selecting a remembered sequence and routing the memory
signal to the reference input of a sequence-control system (see the chapter
in B:CP on memory for more details on this proposition).
When we're talking about things like sunrise, we never do have to develop
an output function for this control system, because the sun always comes up
by itself, and there's no effort we can make that will bring it up sooner
or keep it from coming up. We set the reference signal for sunrise when the
sky gets to a certain degree of lightness, and pretty soon the error goes
to zero.
I don't know if that's really true; it's just my way of trying to explain
the expectancy we feel when sunrise is about to happen.
This does explain, however, why we often _do_ take action when something
that is expected to happen doesn't happen. When you turn the ketchup bottle
upside down, you "expect" ketchup to come out, meaning you _intend_ for
ketchup to come out. Sometimes it does and sometimes it doesn't. If it does
come out you don't have to do anything further, but if it doesn't you
generate a lot of behavior like pounding on the base of the bottle to make
the ketchup come out. In that case, the "expectation" is pretty clearly a
reference signal.
Best,
Bill P.