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

Re: simulation vs. mimesis



Tim,
 
The quote you posted of my father's is divorced from its context, which I have warned time and again is not wise to do.
 
Here is a passage from "Anticipatory Systems" which touches on the issue of what his main contextual base is and how it affects definitions within that context:
 
From AS, page 177:
"In the present chapter, we turn to some specific examples of encoding of Biological systems, with special reference to the way in which characteristically biological qualities are related to the numerical observables manifested by all natural systems. For this reason, we shall not consider reductionistic approaches in detail, nor any of the encodings belonging entirely to biophysics. The essence of such encodings is precisely that they treat biological systems exclusively in terms of their numerical observables. Thus, for example, we shall ignore the large literature on the flow of blood as a hydrodynamic problem, pertaining to the flow of viscous fluid in a family of elastic vessels. For our purposes, this literature is part of hydrodynamics, and not of biology. It is of course true that such studies are often of vital practical importance, but they raise no question of principle beyond what we have already considered in chapter 3.2 above; the biological origins of such studies and their application to cardiovascular problems are essentially irrelevant to the encodings themselves, and the manner in which inferences are drawn from them. We shall concentrate instead on examples of encodings which exhibit a basic metaphorical aspect, and on relational encodings. It is only in these cases that particular biological qualities play a dominant role."
 
Another example of the impact of context on definitions: From "Life, Itself," page 191, Robert Rosen wrote:
"Since the main thrust of the present work is to put organisms on the left-hand side of [the modeling relation] diagram, the limits of formalization are obviously important to us. But as we see, the issue transcends even this and goes to the heart of scientific epistemology itself. The assertion that formalizations suffice in the _expression_ of Natural Law, and hence that causal entailment is to be reflected entirely in algorithms, is a form of Church's Thesis, which I will discuss later... If it were true, the consequences that follow from its truth would clearly have the most staggering implications for all aspects of human thought. For good or ill, however, it is not true, not even in mathematics itself.
 
Let us turn now to the question of what algorithms can actually accomplish in a formal context... [which he begins to do, mathematically]...
 
Conventional parlance has it that any mapping defined in terms of an algorithm, as I have described, is "recursive". Note, however, that this usage is different from my usage of the term in chapter 3. In our parlance, a mapping (on the integers) was recursive if f(n) entailed f(n+1) for every n. In that discussion, I have a necessary and sufficient condition (see 4D.2 above) for such a mapping to be recursive. It is perfectly possible to define mappings f in terms of algorithms, which do not satisfy this condition. Conversely, there is no reason why a mapping recursive in our sense, or rather, the transformation Tf(n)=f(n+1), which generates its values, should be definable in terms of an algorithm. Hence the terms "definable by an algorithm" and "recursive" are not coextensive. Indeed, it will be important for us to carefully differentiate them. Henceforth, I shall call algorithmically definable mappings "simulable," for reasons to become apparent in a moment (they are also variously called computable and effective), and reserve the term recursive for mappings (simulable or not) satisfying [4D.2] above.
 
Thus, the word "simulable" becomes synonymous with "evaluable by a Turing machine." In the picturesque language of Turing machines this means the following: if f is simulable, then there is a Turing machine T such that, for any word w in the domain of f, suitably inscribed on an input tape to T, and for a suitably chosen initial state of T, the machine will halt after a finite number of steps, with f(w) on its output tape...."
 
He goes on to get into more and more detail in this vein, but the point I'm making is that his definitions are contextually linked and he is very careful to make the distinction. When he defines simulation in this way, it is not the same as a model. (He even speaks, on p. 192, of tricking the mapping and the algorithm that computes its values, into evaluating a different map; "This trick is the essence of simulation and of program.") However, he has defined the same words differently in different discussions, in other parts of this same book, as well as other places, and my point is that there is rarely any generalization that can safely be made that ignores context.
 
In this context, he is discussing how simulation pertains to a machine and to programming. He says that the words "model" and "simulation" are not synonyms (in this context) and goes on to explain why. The fact that a simulation in another context can also be a model is what leads to the interchangeability of the two words regardless of context and that's where the problem lies. However, in the context of surrogacy and modeling, there will be different definitions...
 
What I'm trying to get you to see is that in Rosennean Complexity Theory, contextual constraints determine the effect. This is as true of my father's way of writing as it is of natural systems in the universe, according to his body of work (Theory). You've got to be more flexible in your definitions because the definitions change as the context changes. The relation between a good model and the natural system it models is what will remain, and you have gotten that idea down pretty well. But I would caution against getting too caught up in one definition of the difference between simulation and model in a particular context and transpose that definition to other contexts as if it were "a law".
 
Judith

 
----- Original Message -----
From: Tim Gwinn
To: ***
Sent: Friday, December 31, 2004 9:55 AM
Subject: Re: [ROSEN] simulation vs. mimesis

From Rosen's "On the Scope of Syntactics in Mathematics and Science: The Machine Metaphor" in Real Brains, Artificial Minds (Casti/Karlqvist, North-Holland, 1987):
    "The words "modeling" and "simulation" are often used interchangeably as synonyms. In a certain sense, this is justified, since both models and simulations manifest (though in different ways) the features shown in Figure 1 above. [Fig. 1 is a diagram of the Rosen modeling relation - TG] However, as we shall see, the idea of simulation is irrevocably tied to the requirement for an extraneous machine that implements it: something that has no counterpart, or decoding, in the system being simulated. As a result, the causal structure of the system being simulated is not preserved in the simulation. It is this property that distinguishes a simulation from a model, in which causal structure is (or should be) faithfully preserved." [p. 15, ital. orig.]
Regards,
Tim
 
-----Original Message-----
From: ROSEN Forum [mailto:***On Behalf Of Tim Gwinn
Sent: Friday, December 31, 2004 9:14 AM
To: ***
Subject: Re: simulation vs. mimesis

Judith,
 
I can't disagree more. By definition, a modelling relation has the requirement of congruence of entailment structures, which makes the whole diagram commute. If it doesn't, then it is not a modelling relation, no matter how good the correspondence of behavior between the two systems, and the one system is not a model of the other.
 
The entire epistemology rests on making this distinction. Without it, all the arguments in Life Itself collapses.
 
Regards,
Tim
 
-----Original Message-----
From: ROSEN Forum [mailto:***On Behalf Of Judith Rosen
Sent: Friday, December 31, 2004 8:04 AM
To: ***
Subject: Re: simulation vs. mimesis

He (Robert Rosen) also said that anything can be a model if it accurately correlates to aspects in the real system it models. He also said that a system can be a model of another system-- when we use it that way. The answer to the question; "What is a model?" will depend on context. When is a simulation not a model? When you want to understand underlying entailment of the natural system you are purporting to model. When is a simulation also a model? When you are modeling observables. Therefore, the definition of "model" is a definition that is entirely context dependent for each instance of use. You can't argue it in generalities when it comes to a concept like this one. That's what this confusion is being driven by: There are times when a simulation is a model... times when it isn't, ...and times when it is a very poor model (in other words, it is intended to be used as a model but it shouldn't be used as a model because what you need the model to do is not encoded into the model-- it does not commute in the necessary way with the natural system). This third case is what Tim is talking about.
 
You can model more than the entailment of system organization. My father wrote extensively about modeling relations and the reason is because it's a complicated subject and to model something well requires many aspects of human judgment and discernment. It's partly an artform, in that intuition and talent play crucial roles in the process. You have to know what you need the model to do before you begin and trouble can start even with those decisions. This is the case with using simulation to try and learn about underlying entailment relations.
 
Judith