[Date Prev][Date Next][Thread Prev][Thread Next]   [Date Index] [Thread Index] [Author Index

Re: Rosen & Ashby --> Syntactic vs. semantic limitations




Hi Pete,
 
I really like your phrasing:
[PVG]
"In my terminology, we create the syntactic  structure that defines the relationships among the semantic referents.".
 
I have a somewhat different take on the N-body system problem. It comes mainly from Rosen's mentions of inertial vs gravitational aspects in chapter 1 and p. 109 of Essays.
 
In the Newtonian view, where fundamental particles are featureless, if we had 3 particles as the 3 referents, being pushed around by forces from the environment, then syntax alone would be entirely adequate to define the dynamical relationships among the referents. And not only would it be syntax, but computable syntax (algorithm) at that. And so the system would be simple. Even if the particles had certain qualities, such as inertial mass, which do not depend on anything else about the system, the system would remain simple. So, just having 3 or more semantic referents is, in my view, not sufficient to cause a system to be complex. Newtonian mechanics readily deals with large numbers of particles with inertial mass.
 
But once particles are endowed with the ability to generate forces, such as gravity, then the relationship between such particles is no longer purely syntactic. Instead, the system now has, aside from the referents to the bodies, additional referents in the form of dependencies of each relationship upon the other relationships in the system. So, one relationship refers to another relationship which, in turn, refers back to the first relationship. Voila - classic loops of reference.
 
In a 2-body system, there is only one relationship between the two bodies, and hence, no loop of reference. But for N-body systems, where N>2, the loops of reference, or impredicativities, cannot be avoided. And so, the systems are complex. And the system fails to be modelable in a purely predicative syntactic form, which is required for a model to be algorithmic. As such, these systems possess noncomputable models. 
 
It also becomes immediately clear why such systems cannot be solved reductionistically. Attempting to do so breaks the loops of reference in the very act of partitioning, say, a 3-body system into a 2-body and a 1-body system. So, the problems solved for the separate 2- and 1-body cases cannot address the problem of the loops that are found only in the whole 3-body system.
 
 
With regard to:
[PVG]
Granted, I picked a simple system to make it easy to illustrate the point. Had I picked a three-body system rather than a two-body system, the syntax would have been quite a bit more complicated, requiring me to treat two of the bodies -- say, m1 and m2 -- as a single, composite "virtual mass" that interacts with m3 , and then consider m2 and m3 as a composite virtual mass that interacts with m1, and so on... and then run the calculations iteratively to generate my results within certain specified ranges of values for the variables. It's more complicated, but there is still no indeterminacy in the syntax.
The procedure described is, of course, an example of a simulation, and is not a model. This can be easily seen by trying to put the inferential entailment structure of such a procedure into congruence with the causal entailment structure of an 3-body system. In the 3-body system, all 3 bodies move simultaneously - the causal forces apply to each other simultaneously. This is entirely different than the stepwise grouping of two masses, which then entails a total force on the third mass, and then running it for the other two groupings. Such simulations are useful, but they only mimic behavior of the system.  They do not embody the same entailment structure as a model would. So, lack of precision is a pragmatic issue as far as successful mimicry of behavior is concerned, but precision itself is not a criterion for deciding simplicity vs. complexity or model vs. simulation. 
 
I believe it would be possible to create a purely syntactic, but necessarily noncomputable (because of those loops), model of the 3-body relationships (or even better, for a generalized N-body system) using something along the lines of a relational model, as exemplified by the (M,R)-system.
 
Regards,
Tim
 
 
-----Original Message-----
From: ROSEN Forum [mailto:***On Behalf Of Pete Giansante
Sent: Saturday, November 22, 2003 5:13 AM
To: ***
Subject: Re: Rosen & Ashby --> Syntactic vs. semantic limitations

Hi John K., Tim, et al:

  ----snip-----
 
 
 ~As RR (via Tim) correctly points out, we impute the relations to the qualities we observe. In my terminology, we create the syntactic  structure that defines the relationships among the semantic referents. I agree with RR's assertion that the relations (the "syntactic relationships") are mental constructs. Since the "syntactic relationships" we create are constructs of the human mind, we can define them to any degree of precision we wish, and philosophers & theologians are unabashed in doing just that. In science, however, we're accountable to more sharply defined constraints; our [syntactic relationships among semantic referents] = [models] are subject to empirical corroboration. The requirement for corroboration in the "material world" is universal in any discipline that purports to be a "science".

In general, syntax is not the principal epistemological limitation on the precision (and therefore the functionality) of our models. The history of science is replete with examples of apparent roadblocks in our ability to organize percepts/semantic referents into a coherent syntactic structure that defines the nature of their relationship to each other. In more familiar terminology, nature frequently stumps us in our attempt to create theoretical models. Yet, time and time again, we develop ever more ingenious syntactic statements that sidestep such apparent obstacles -- the "mathematical tricks" that allow us to bypass the roadblocks -- with the result that new theoretical models emerge... models that ultimately prove to be empirically corroborable.

Neither science nor philosophy knows "why" the capability of the human mind to think mathematically turns out to be reflected in the syntactic relationships among the semantic referents we perceive -- or as some would say, "in the structure of the universe" -- but the fact that the human mind has that capability is indisputable. For the sake of staying on point, I'd prefer to avoid the philosophical implications of that congruence in this discussion, except to say that it seems to me to be a remarkable coincidence. As Paul Davies has observed, it almost seems fishy. 'Nuff said.

What we've learned from even our most successful physical theories is that the epistemological limitations on the precision/functionality of our models are principally imposed by semantic constraints, not syntactic ones. That is, limitations in our knowledge structures derive from our inability to specify or describe "the things we are talking about" -- i.e., the semantic referents themselves -- with infinite precision, not necessarily the relationships between those things.

For example, we have no difficulty with syntactic imprecision in the following statement:
F = G • m1• m2 / r2
... the well-known inverse square relationship for the gravitational force of attraction in a two-body system. When we make predictions using that statement (in a classical, non-relativistic context) , those predictions can be robustly corroborated by observations that fit our predictions to any degree of precision of which our measuring instruments are capable.

But measurement is a semantic process, not a syntactic one; that is, measurement is the process by which we define the quantities represented by the variables in the inverse square relation. It's the process by which we associate symbols with their referents, and therein lies the limitation on precision. We cannot measure the masses or their separation distances to infinite precision. Hence, it is our inability to specify or define those semantic referents that imposes the limitations on precision, not the syntactic structure that defines the relationship between the semantic elements.

The syntactic elements in the inverse square statement above are the mathematical symbols:
  • the "=" sign;
  • the scalar multiplication operators "•" between G, m1, and m2; and,
  • the division operator " / ".
The operations specified by those symbols are precisely defined with zero uncertainty, hence there is zero non-specificity in those syntactic elements.

Granted, I picked a simple system to make it easy to illustrate the point. Had I picked a three-body system rather than a two-body system, the syntax would have been quite a bit more complicated, requiring me to treat two of the bodies -- say, m1 and m2 -- as a single, composite "virtual mass" that interacts with m3 , and then consider m2 and m3 as a composite virtual mass that interacts with m1, and so on... and then run the calculations iteratively to generate my results within certain specified ranges of values for the variables. It's more complicated, but there is still no indeterminacy in the syntax.

But is it a fair example, since I've picked a relatively simple system -- even with the three-body system? Yes, it's fair... because JJK's statement was unequivocal:
...there is no absolutely precise syntactic statement of anything (e.g., Rosen's theory) that can be made.
I'm assuming that the statement was intended to apply universally, irrespective of whether any given "syntactic statement" is being made about  a simple system or a complex system.

What may be surprising is that the three-body system is NOT a simple system -- at least not in the sense that it is simulable. We cannot create an algorithmic representation of the system that is a more concise model of the system, which model enables us to predict the future states of the system with greater accuracy than simply letting the system operate and observing those future states in real time.

This is not a particularly recent revelation. As Poincaré demonstrated a century ago, the system is non-integrable, a characteristic that Prigogine has since shown is the basis of the complexity that engenders "life itself." (His words!) The dynamics of the system are complex in that they are increasingly unpredictable with increasing ∆t (system run-time) relative to any given starting time t0, because of an essentially infinite sensitivity to the precise values of the initial conditions. How would we determine those values? By measuring them, and we can't measure them to infinite precision. So all of the imprecision in our ability to define the system state at any given future time t is, once again, a result of a limitation on semantic precision, not on syntactic precision.

That's not to say that we are able to cough up the same high degree of syntactic precision that we can attain in simple systems in our statements about systems that are vastly more complex, even if their complexity arises only because they comprise a much higher number of elements, components, sub-systems, etc. There are obviously some very significant challenges involved in modeling such complex systems, as RR so convincingly observed. Many of the syntactic "tricks" that have been so successful in the domain of relatively simple physical systems are utterly powerless to model the dynamics of even the most primitive self-organizing systems, let alone the vastly more complex, principally self-referential ones.

It remains a mystery to me how anyone can deny the overwhelming inapplicability of context-independent, principally syntactic models as adequate descriptions of complex systems. I happen to place a great deal of weight on Gödel's Incompleteness Theorem as an epistemological limitation on the sufficiency of purely syntactic models to describe all natural systems, but whether or not one shares my appreciation for the profound implications of Gödel's work is really quite beside the point. As I've demonstrated with the example above, even where an absolutely precise syntactic relationship can be established as a model of system  dynamics, there is already a fundamental semantic limitation in the model, which limitation renders the system complex. Essentially, all it takes to enable the transition from simple natural dynamical systems to complex natural dynamical systems is the multiplicity of system elements (i.e., semantic referents) beyond the number 2.

If all it takes to engender complexity is 3 semantic elements, RR's conclusion that simple systems are the non-generic ones is inescapable.

Best regards,

PVG
© 2003 Peter V. Giansante -- All rights reserved.