At 09:59 AM 7/23/98 -0400, you wrote:
>More comments on metaphors and models
>
>John:>>>I wrote in the autevolution paper on the problem with metaphore. Yes,
>>>>everything may be metaphorical at some level, but we do make a distinction
>>>>between the epistemology of science and more colloquial metaphores.
>>>
>Alexei:>>This distinction is quantitative, not qualitative.
>>
>John:>I don't understand this comment. The distinction in epistemology is
>>"mechanism" (which is an unfortunate term now that we are discovering a
>>non-mechanical reality). Metaphores are statements of similarity without
>>any proposed means for explaining the "cause" of that similarity. There's
>>no need to get too esoteric about "mechanism" and "cause" here because
>>there are plenty of simple examples. Darwin proposed "survival of the
>>fittest" (reluctantly adopting these words, actually) as a metaphore, and
>>said so. After the "modern synthesis," which provided the "mechanism" for
>>evolution it was no longer considered a metaphore.
>
>This is an important issue because it draws a line between realists (John)
>and pragmatists (me). Realists say that metaphors imitate reality
>wheras models represent reality (capture "causes" and "mechanisms"). This
>view in well supported by scientific practice where we can see both
>phenomenological metaphors (or "instrumental models") and mechanistic
>models that attempt to represent real proscesses. Although I agree that
>these categories of models exist, I don't see a qualitative difference
>between them.
The labels "realist," "pragmatist," or any labels are stereotypes that
attempt to identify one with a past belief rather than recognizing
important or slight differences in present interpretation, which is the
point of having a conversation. Our current views are a mixture of these
past categories, each of us giving emphasis to elements we think are
important. The philosophy of science is evolving, and hopefully these
discussions add constructively while reminding us of certain aspects of
past epistemologies that have value and should not be lost. I doubt that I
could qualify in any category, even as a good "realist," or that Alexei is
a rigid "pragmatist," but perhaps we have each learned something useful in
this area. Still, all we can do is retell the stories in ways that extract
various insights.
My short answer to this point about a qualitative difference is that the
difference is not in the character of the model or explanation itself - it
is in the goal and method of the exercise, and in the methphysical belief
in a singular or multiple reality. "Science" is not defined as a collection
of models, although it produces them -- it is defined as a method of
discovery (emphasis on "a" because there is another method too - experience
and introspection which is more associated with metaphores). Metaphores do
not have explanation as a goal -- they don't need consistent explanation:
they're true because they exist. That is qualitatively different from
having explanation as a goal and as a test. Also, the method of metaphor is
comparative experience or description based on intuition or belief ("love
is a rose," "mother nature," the machine metaphor, Darwin's "stuggle for
survival," Ptolemy's solar system model, etc.); whereas the "scientific
method" involves experimental testing or analysis to attempt to discredit
the metaphor -- to show where it is untrue, and therefore not universal,
rather than to seek out those specific cases where it does seem to work, as
one intentionally does with a metaphor, ignoring the exceptions. So I think
both goal and method are qualitatively different, and very importantly so.
Here's the long-winded version of how I write the story to support this
answer:
diminising number that we might put in that category today) spoke of causal
explanations as SEEKING to describe something "real" regarding causation,
which they ALSO though might lend itself to a single description. Today we
mostly keep the first idea but are relaxing the constraint of a single
description. Rejecting all of it would be "throwing the baby out with the
bathwater." The old idea tended to elevate one's model of reality to the
status of reality itself -- but this was mostly in the lay person's mind
even though it may have also been a strong lure for many scientists who
liked to think they were unraveling the "secrets of the universe" rather
than just playing with models (they were mostly men with certain
patriarchal ideas, and "playing with models" would be something they would
associate more with girls at the time).
Most philosophers I've read point out that much of this fixation of model
to reality is socially driven - not in fact the intent of the more
enlightened scientists, many of whom recognized that no model can be a
complete description of reality. In fact, that idea is fundamental to the
hypothetico-deductive method itself, even though this "positivist" approach
(H-D methods) is now taken as an example of rigid and limited dogmatisim
because it only deals with experimental procedure. H-D methods presume up
front, however, that nothing can be fully known conclusively, hence the
emphasis on disproof (by experiment) rather than proof -- the idea was that
we can never say our model of reality is correct, we can only look for its
flaws and in their absence gain confidence. Physicists generally seem to
understand this (except perhaps the one's trying to create a "theory of
everything" -- although I suspect that even they know better and are just
playing to the media). However, the positivist approach was too rigid
because it lacked a means for shifting the whole basis, or paradigm, for
theory construction, i.e., "revolutions." Because all theory depends on
one's beginning assumptions and metaphysics, it is possible to develop good
theories (and bad ones) from different starting points, and each starting
point has its merits and demerits. The point, which I describe elsewhere in
an extensive analysis of epistemology
(http://www.bayside.net/NPO/BMI/autevol/ghw_epi.htm), is that H-D methods
work very well within a given worldview (or paradigm), but if pursued
rigorously will eventually exhaust the explanatory possibilities of that
view. I'll skip the examples here. The result will be a scientific "crisis"
and "revolution" as discussed by T. Kuhn (but Kuhn did not describe in
detail how it works, as I attempt in the paper).
The scientific revolution involves a shift in perspective, worldview,
assumptions, or paradigm -- but this is not a random shift. There are most
often definite consistencies that are maintained when switching paradigms.
Einstein, for example, described this process extremely well in his 1916
lay person's guide to how he developed the theory of relativity. Even
though "reality" can be seen from many different perspectives, what we
learn from those perspectives must be consistent between views -- that is
if we maintain the belief in a singular reality (as opposed to a singular
description) and thus a basis for SEEKING synthesis and integration (i.e.,
the ultimate goal of interrelating theory elements to show consistency
between theories). This is a very key distinction -- the theories
themselves may not be interchangeable or derivable from each other, or
amalgamated into one "super theory," but what they describe and predict
must be consistent, otherwise we are altering the basic metaphysic of a
unitary reality. If you do that, i.e., adopt the view that there are
"multiple realities," and therefore believe that alternate theories do not
need to offer consistent results, you are becoming a "realist" of a much
worse variety -- one who believes in two or more realities! Let me support
this statement philosophically: Neither view, single vs. multiple reality,
can be tested scientifically. The choice is therefore metaphysical. If,
however, a singular reality is correct, knowing about it would certainly
constitute greater knowledge than knowing only the separate views and not
their connection. Science seeks knowledge, therefore it must SEEK to
understand nature as a singular reality, even if it only finds multiple
aspects (as represented in current multiple theories).
> The only difference is in the number of hierarchical levels
>they try to represent. Phenomenological models do not go beyond 1 level
>(i.e., the surface), and mechanistic models go deeper, representing
>2, 3, or more hierarchical levels. Different phenomena tend to have
>common deep hierarchical layers. Thus mechanistic models are more
>integrated with each other, whereas phenomenological (instrumental)
>models do not form a system. Mechanistic models are more stable historically
>because they form a coherent system.
Again, I would say this is HOW IT LOOKS. But the qualitative difference is
regarding WHAT IT SEEKS. The distinction is in the method and goal, not
strictly in the construction of the model. The phenomenological model,
instrumental model, and metaphore do not seek explanation at deeper levels
or interconnection between phenomenaological models (unless connections
appear at the phenomenaological level). They simply don't look deeper for
understanding and as a consequence, don't look rigorously for where they
are wrong. Where they are wrong, they just don't apply because it is
presumed to be a separate reality. Reality in this view is phenomenological
(and fragmented). They don't seek generality and consequently can't find it.
>It is like a crossword puzzle:
>the more intersection of words you have, the smaller is the number
>of ways to fill the puzzle. But there is no guarantee that the solution
>of the puzzle that you have found is the only possible solution. May be
>it is the only possible solution for the current English language (as
>it is represented by Webster dictionary), but languages evolve, and it
>may happen that in the future we will get a second solution for the same
>crossword puzzle.
>
Yes, this is a good, but perhaps a more appropriate analogy would be if we
are given the crossword solutions and, as in science, must find the rules.
Different "solutions" represent different patterns we might observe, but we
seek to understand what rules they have in common and which ones are
unique. The models might get at a set of rules that explain one pattern but
not the other, and we might decide they have completely separate sets of
rules. But if we seek common explanations a useful connection between the
two patterns may become known.
An extremely frustrating aspect of ecology right now is the tremendous
number of phenomenological theories that each start out from a different
basis and make no connection with each other. It has fueled century long
debates (e.g., the Gleason/Clements community debate) with little increase
in understanding. To put it differently, each camp promotes their own
syntax and we forget to look for a common meaning. If we get a second set
of questions for the two crossword patterns, did we solve two separate
puzzles, or do they have something in common? Science should be obligated
to investigate this question.
-----------------------------------------------
John J. Kineman, Physical Scientist/Ecologist
National Geophysical Data Center
325 Broadway E/GC1 (3100 Marine St. Rm: A-152)
Boulder, Colorado 80303 USA
(303) 497-6900 (phone)
(303) 497-6513 (fax)
jjk@ngdc.noaa.gov (email)