Shannon's comment -Reply

[Hans Blom, 951020]

(Bruce Abbott (951019.1540 EST))

The research process begins with observation. These observations
reveal phenomena that require explanation. To develop a scientific
explanation, one must begin by framing hypotheses that can be
subjected to objective test. ... This leads to the application of
the Test for the controlled variable, to confirm or deny this
hypothesis.

Let me qualify this somewhat.

The research process begins with observation.

I doubt that this is where it begins. We "observe" all day, because
we're happily controlling away all the time in a mostly expert
fashion. We're experts in doing the things that we do.

But once in a while we experience a "startle", when something isn't
quite the way we thought it would be or "ought to" be. Something
strange happens, something was not anticipated, we have no expla-
nation for it, and that kind of shocks us. This emotional startle
response is, I think, what begins an investigation / research
process.

In terms of model-based control: the prediction by the internal
world-model is noticeably (startlingly) incorrect.

Now, in many cases people don't care very much for a process of
"scientific" investigation, maybe because the investigative effort to
"explain" (i.e. model) what happened would be too much and gain too
little. This leaves the person with some doubt, uncertainty or
confusion. But a little of that doesn't hamper efficient control.

These observations reveal phenomena that require explanation.

Shockingly discrepant observations require explanation. But the funny
thing is that the "explanation" will always be in terms of the al-
ready existing internal world-model, in such a way that is minimally
modified. That is why peoples' "explanations" depend, to a large
extent, on their personal histories. That goes for scientists, too.
If you're educated in S-R theory, you will have a tendency to explain
new findings in the light of this theory -- and make small extentions
or changes to it. If you're educated as a control engineer, you'll
tend to explain things in terms of control theory. If you happen to
know model-based control theory, you tend to explain things from
_that_ perspective. This is a kind of conservatism that is inherent
in all systems that build models. Throwing away the model is the same
as throwing away all previously acquired knowledge, which very likely
results in total loss of control. We just cannot afford that. If we
happen to loose control over an important aspect of our lives -- in a
situation of personal crisis when something happens that we hadn't
thought would be possible and that we have no experience in at all --
we hope and try to make sure that something like that will never
happen again. Why expect such courage in scientists?

To develop a scientific explanation, one must begin by framing
hypotheses that can be subjected to objective test.

I maintain that most people don't do this, at least not habitually.
Moreover, which hypotheses _can_ you frame? Only such as your current
knowledge _could_ generate. And how this is done is still a mystery.
It is not a purely logical process, because the number of possible
hypotheses might be extremely large. Introspection seems to indicate
that usually previous pieces of knowledge suddenly snap together and
a kind of deeply felt Aha-Erlebnis results: "yes, _that's_ the solu-
tion!", with a concomitant significant -- but subjective -- feeling
of certainty. But an _explanation_ for such "conversion experiences"
is still lacking.

In the daily practice of science, things aren't too different. Most
research is started from a hunch: "I think that this might be true,
let's try to demonstrate it". None of that Popper-like trying to
contradict your own (subjectively certain) theories.

What you describe is a kind of glorified theory that doesn't have
much to do with the real world. Medical students are taught the
procedure of "differential diagnosis": given the signs and symptoms
of the patient, generate all possible hypotheses that could explain
those signs and symptoms in terms of a disease, and subsequently try
to (dis)prove all disease hypotheses until one is left. That is then
the true diagnosis.

This process doesn't work as well as it should. For one thing, the
number of diagnoses generated is often insufficient in that it does
not include the true diagnosis. As you might have thought, this is
usually the case with rare diseases that the physician isn't familiar
with. However rigorously The Test is executed, it doesn't lead to the
correct result if the correct result isn't even considered as one of
the possible outcomes. Second, the differential diagnosis process is
far too time-consuming to have practical value in everyday life. But
in medicine, it is still the option of last resort when everything
else has failed.

... This leads to the application of the Test for the controlled
variable, to confirm or deny this hypothesis.

So this is something of extraordinary rarity. Even in science. Even
in PCT. Where are the hypotheses? Currently, PCT seems mostly con-
cerned with educating people about the concept of control -- and a
(too) limited concept, at that, in my opinion. Where are the hypo-
theses about what people control? The researcher that has come
closest in this regard was, I think, Maslow, who made lists of "basic
needs". Thus far, I haven't seen a PCT version of such a list. Yet,
in my opinion, that would be essential. Given a list of hypotheses,
one might start to (dis)prove its items. PCT hasn't even started to
do this...

So what you describe is an ideal much more than practice, even in
these here PCT circles...

Greetings,

Hans