Re: Consistency vs. Correspondence

["John Kineman" <***>]

Tim Gwinn wrote:

Hi John,

 

That helps quite a bit, and was very informative.

 

Regarding 'consistency', I am inclined to say that "consistency is when both theories commute in their respective modeling relations with the same natural system (in the same context)".  How does that sound?

Yes, I think so. In the same context (scale, precision, etc.) is essential.

I placed "in the same context" in parentheses, because I am not sure if that is the best phrase.  It seems appropriate where, for example, classical and relativistic theories are consistent, but only within certain limited contexts of gravity, velocities, etc. But I also wonder if there might be cases where the theory itself may, for example, dictate certain measurement processes that differ from that of another theory, such that the theories themselves alter the context of the system under study.

Perhaps if we clarify it as "epistemological context." That excludes the
causal (what I perhaps controversally called "ontological" before) feedback
from model to system.

In the latter case, consistency may not occur, due to a change in context that is unavoidably due to the theory itself. (No specific instances come to mind, it just occurred to me as a possibility that might arise when comparing theories.)

Ah! The specific instances are organisms. Rosen is clear in AS which: "Presents
the first detailed study of this most important class of systems which contain
internal predicative models of themselves and/or of their environments and
whose predictions are utilized for the purposes of present control." (from
the book cover!!!).

So, a mechanical model of an organism would predict events within uncertainty
limits, which is consistent with a more exact knowledge of intentions on
the part of the organism (say human acting in society). There would be no
disagreement in saying "Joe visits the East End Tavern on the average of
5 times a week" even though Joe can change his habits and also can add all
kinds of different reasons for visiting the bar which the mechanical observation
cannot represent.

 

 

Rambling on a related topic......the thing that has always troubled me about instrumentalism is that it has always seemed unclear to me on what basis a theory's results are supposed to be "extendable" to other situations. Instrumentalism (as I understand it)  takes a theory to be neither true nor false, but instead as just "useful" or not to some degree. What "usefulness" exactly means varies, I gather, with the particular flavor of instrumentalism, but it seems to generally be around the notion of the ability to make "good" predictions that match the phenomena. 

Yes, this troubled me too and I resisted the idea for quite a while. I began
as a realist (in physics). But I think there is good and bad instrumentalism.
The bad takes the excuse to propose anything on any basis and make no claim
at all about generality. That is nothing but a description to me. But, by
allowing for multiple relative reality views, we do not have to abandon principles
of generality, necessity, parsimony, consistency, and formalizability, and
usefulness. (I describe these six principles on my website at http://www.nexial.org/bmi/autevol/ghw_epi.htm).
So, if each theory seeks generality (what I also refer to as relative realism),
is formulated in an efficient or even elegant manner, is consistent with
results of other theories, responds to a need and is useful; then it is good
instrumentalism or relative realism. There are all kinds of theories that
would fail this test.

 

If theory X is "useful" as a theory in situation Y, then what is it that allows or suggests that theory X can make any assertions about situation Z?

In realism, the illusion that we discovered a "real" Platonic rule for nature.
In instrumentalism, only testing would be convincing. But there is another
process that Rosen discusses, which is "reasoning by analogy."  He wrote
(AS, pg. 400): "Conversely, we found that many different natural systems
can admit the same model; this provided the basis for the fundamental concept
of analogy between systems, and the use of analogy as a powerful scientific
tool." So, a partial answer is analogy, but certainly testing is also needed.

And further, if theory X is "useful" as a theory in situation Y, then what is it that allows or suggests that some logical implication(s) of theory X can make any assertions about situation Z? This has always puzzled me - if a theory is just "useful", then what gives it any force of argument beyond the specific situation(s) in which it is "useful"? And even within the situations for which it is useful, does it have any force of argument?

 

Again, in realism the force of argument was that we discovered Platonic reality
and are merely testing and refining that discovery, hence everything should
comply with our wishes. But that was an illusion, and a damaging one in addition.
I think instrumentalism recognizes the problem of the modeling relation -
that there are an infinite number of ways to relate to or model a natural
system. So then the epistemological rules (six principles above), and reasoning
by analogy are the best tools available to ensure progress in knowledge,
recognizing that it will always be approximate and incomplete. This was VERY
dissatisfying to the physicists who were forced to accept it after Einstein,
and particularly dissatisfying to Einstein who spent the remainder of his
life trying to disprove it (the famous "God doesn't play dice" quote). But
I really think Rosen begins to put reality back together again, if we can
cautiously give (relative) reality to the modeling relation itself. i.e.,
it is a better "reality" than we had. This is why I truly believe Rosen was
a genious on a par with Newton,  Einstein, Darwin, and other giants in their
contexts. It may be 100 years before that is generally recognized, or perhaps
necessity will bring it sooner.

In the cases above, particularly in the latter one about the logical implications of a theory being applicable elsewhere, this strikes me as a tacit assertion of what Rosen makes explicit: that inference in a theory and causality in the world are being related to each other. In the modeling relation approach, we say - when the MR commutes - that the causal structure of the natural system is congruent with the inferential structure in the model. This relationship and congruence of entailment structures also provides a basis for how it is that implications of theories can be proposed to say something about some other situation. In instrumentalism, I am unclear how, or if, that relationship (and its consequences for extending a theory to new situations) is codified.

I think you are right, regarding the common way instrumentalism is imagined
(if anyone is really cognizant of being an instrumentalist anyway, or if
this is just a label made by philosophers). Rosen makes it reasonable and
at the same time shows why realism won't work. The modeling relation is the
answer to both questions.

There is another subtle aspect that relates to this question. It is the hierarchical
nesting of systems that I tried to explain in earlier posts. Set aside doubt
for a moment as to whether this was Rosen's intention (I believe I can show
it was, but suspend that for further discussion). If such nesting is the
case, then a model is not isolated to any single modeling relation. As Rosen
states in AS (i misplaced the page reference, but was reading last night),
this results in "an infinite mathematical object." The model itself (any
one we choose to identify) also exists in relationships with larger and smaller
systems which contain models (again this mixes the epistemological and ontological,
which Rosen continuously did, unfortunately without always being clear when
changing gears). This also gives some basis for saying models should be more
generally applicable than to "totally isolated cases," because there is no
such thing as a totally isolated case. So, if you come up with a good explanation
of Joe going to the bar, chances are there will be other situations that
are analogous, because that formal behavior itself does not occur in isolation
from the universe.

 

This (congruence of entailments) as a requirement for being "useful" seems to be something that instrumentalism wants to disavow. 

I'm sure many will be sloppy and mis-apply instrumentalism, which is why
I was adverse to it earlier. If not carefully considered, it can be an excuse
for being arbitrary. That is why I did some looking into the epistemological
"rules" cited above and considered how they would apply. 

By the way, you will notice that "falisfiability" is not in my six rules.
That is because it doesn't apply to world views or meta-theories (i.e., the
basis for theory, as in Rosen taken as a new outlook on nature vs the mechanistic
assumptions). With regard to specific theory predictions within an established
view, add falsifiability as well.

Yet, it would seem that it must also (tacitly, at least) employ them if one is to engage in arguing or proposing that if a theory leads logically to some conclusion, then that also argues that the resulting implication structure will be mirrored in the causal reality it describes.

I think good instrumental theory will seek generality and look for ways of
integrating or at least relating its theory structure to other theories.
Otherwise we get a tower of babble. Necessity is an important consideration
which I spent some time on in the web page cited above. As a general rule,
one should reach the point of inventing a new instrumental view of reality
only after identifying a clear paradox in other prior views and then proposing
a new one that resolves that crisis. That is one very powerful check on arbitrary
constructions, although it shouldn't preclude the occasional leap of fancy,
which then should be followed up with justification in these terms.

So, I guess I advocate (and think Rosen is exemplary of) a relative realism
and provisional instrumentalism; certainly avoiding the extremes of both.

 

Just wondering out loud...

As are we all!

Cheers,

J. Kineman