Grush mush

[from Mary Powers (971211)]

The dissertation by Rick Grush that Rupert found is at
www.artsci.wustl.edu/rgrush/--if anyone wants to take a look. The committee
for Grush was headed by Paul Churchland, and included Patricia Churchland,
some people we don't recognize, and a gent who appears to be an engineer,
teaching neurocomputing and electronic and computer engineering at UCSD.

One truly dismaying fact is that all these people bought into
this paper and gave the guy his doctorate. It is based on model-based
control theory, and is an incoherent account of that. It shows no grasp
whatsoever of control theory basics.

It gives Bill the feeling that all the while he has been trying to build a
bridge between control theory and psychology, and focussing on the gap
between them that continues to persist, a merry crew has been ripping the
bridge up behind him. Apparently Hans is not the only one to consider
Bill's version of control theory obsolete and old-fashioned.

What I think has happened is this. Control theory did indeed have its
difficulties in the early years, giving rise to the various myths like
"feedback is too slow", "control systems oscillate", etc. One way to solve
them was to go to model-based control. This made sense to people who didn't
have much feel for control systems, and who were primarily focussed on
getting the control system to produce the right output in a controlled
situation. The other approach was to work within the control system,
adjusting various parameters (such as loop gain) and tweaking it until it
worked. This requires a "feeling for the mechanism", which Bill has said
feels almost like cheating, because it works so well. He basically puts
himself in the system's place and looks for what is wrong from that
perspective. This of course makes perfect sense if you are designing and
building control systems and at the same time figuring out how living
control systems work, including yourself.

It would seem, from what we have heard (from a course our son took on
control systems, and from a couple of engineering teachers who have heard
Bill talk at conferences, etc.) that basic control theory, the simple,
fundamental organization underlying PCT, is barely touched on in engineering
courses. Engineers come away with no real sense of how a control system
works, but with lots of technique - for building emulators and model-based
control systems. The notion of computed output is apparently deeply
entrenched in engineering, no matter how many feedback loops their designs
include, and cog sci people simply accept it, since it conforms to their
preconceived ideas as well.

Mary P.

[From Bruce Gregory (971211.1710 EST)]

Mary Powers (971211)

It would seem, from what we have heard (from a course our son took on
control systems, and from a couple of engineering teachers who have heard
Bill talk at conferences, etc.) that basic control theory, the simple,
fundamental organization underlying PCT, is barely touched on in engineering
courses. Engineers come away with no real sense of how a control system
works, but with lots of technique - for building emulators and model-based
control systems. The notion of computed output is apparently deeply
entrenched in engineering, no matter how many feedback loops their designs
include, and cog sci people simply accept it, since it conforms to their
preconceived ideas as well.

Your assessment unfortunately matches my own, although you have
more data that I. The metaphor of mind/brain as computational
device has become so prevalent that the fact that it _is_ a
metaphor is totally lost on entire communities. A cautionary
tale that, unfortunately, few have had the opportunity to learn
from.

Bruce