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

Re: The notion of semantics in Rosen's writings



Hi Carlos,
 
Good questions. I will do my best to help. I think the clearest way to handle it is to insert explanations into the body of your text:
 
Carlos L. posted:
RR wrote: "An essential part of any language resides in its
implication structure. This comprises a system of
entailments, which generally are expressed in the
terse form P [entails] Q. The main property of such
implications, or entailments, is that they propagate
"truth" hereditarily across them; thus if P is assumed
to be "true" (whatever truth may be, here), then Q
must be true also."
 
This is the standard modus operandi of theoretical science, as a matter of fact. "If P is true, then that means Q must be true." In mathematics (which is another language, with its own implication structure, different from English) it's written as an arrow  ( P ---> Q ). Another way to look at it is; "Q depends on the truth of P, therefore in order for Q to be true, P must be true."
 
An example might be: "If we have a clear sky tonight, we'll be able to see the Perseid meteor shower" (which is still continuing, by the way). So, P = "clear sky tonight" and Q = "see the Perseid meteor shower". If P is not true, then Q will not be true, either: if the sky is not clear, we will not be able to see the stars.
 
This goes in the other direction, too. If someone says; "I was watching the shooting stars last night-- it was AMAZING!" You know that the sky had to be clear.
 
C.L:
RR wrote: "It is often supposed that such
inferential entailments are part of the syntax of the
language and do not themselves depend on any external
referents or meanings; only the truth values arise
semantically. Of course, such truth values may simply
be posited, but in either case, they come from outside
the syntax itself."
 
The fact that we can give examples of this kind of entailment structure (If P, then Q) in both English and in mathematics actually constitutes a proof of both the statement above, and also an example of a modeling relation. 
 
As a proof that the truth values come outside the syntax, itself: entailment is the basis of all causality. That's something that is true outside of all human languages. In a sense, entailment is P and causality is Q. The fact that entailment patterns hold, regardless of the venue is the only reason we can "do science" at all, because this fact is what allows us to make models that actually can predict the future behavior of the system being modeled. 
 
CL wrote: Then he proposes something called 'modeling relation'.
In this 'modeling relation', he says, there is
'encoding' and 'decoding' between what he calls (at
least, in the book 'Life Itself') as 'natural system'
and 'formal system'. Rosen says that a 'natural
system' has a set of 'causal entailments' and a
'formal system' has a set of 'inferential
entailments'.
 
Right. So, assuming that the system we want to model is a natural system, the entailment structure of the system is called "causal entailment" and the entailment structure of the model is called "inferential entailment". So, in our "If P, then Q" scenario... that can be a model if the entailments are accurately transferred from the natural system we want to model. Let's take another system and try to model it:
 
Let's say we want to model the causality of frost forming on our car, as it sits outside in the driveway in mid-October in New York. In autumn, around here, the overnight temperatures are low enough to cause a heavy dew to form on all surfaces outdoors. If the temperatures go below freezing, the dew crystallizes on all those surfaces. It looks really pretty, actually! But it's a real pain if you have to drive somewhere early in the morning: You have to scrape it off all your car windows or else you can't see to drive. So, to form our model, using this mathematical language: P = "temperatures below freezing" and Q = "frost on car windows".
 
We've just encoded the entailment, from the natural system into our model, by our choices of what aspects of the model represent which aspects of the actual system. Then, to test it, we have to decode: Our model predicts several things. It predicts that frost will not be able to form unless the temperatures go below freezing, and it predicts that the freezing temperatures must precede the formation of frost. They are not simultaneous, in other words.  So, from our model, we can predict that if there is frost forming on the car, the temperatures must already be below freezing. Do you see? Making predictions about the real system's behavior, based on our model's entailment structure, and then conducting a series of experiments to see whether our predictions hold true, we are "decoding"--  following the lines of entailment in our model to create predictions and then applying those predicted entailments back to our actual system. If the predictions prove to be accurate, then our model "commutes".
 

The first problem I have is with the definition of
'encoding' and 'decoding'. Rosen defines in the book
Life Itself encoding and decoding as the 'translation'
between 'formalisms'.
 
That's only one particular kind of modeling relation. He begins with the relation between a natural system and a model of it. Then he goes on to say that we can make models of models, too. Any model is a type of "formalism". So, if we have a model and we want to create another model: another way of studying the entailment structure that the first model possesses... then we must go through the process of encoding that entailment structure from the first model into the second model, and also the process of decoding by making predictions, using the new model's entailments to predict what the original model's entailments would predict. An example of this would be to take a statistical model and create a graph from it. The graph is only an accurate representation if the information from the statistical model is converted accurately, from one form into the other form. All the entailment relations contained in the statistical model must be transferred over. That's encoding. See?
 
CL wrote: Also, is not very clear to me what qualifies as a
'natural system', since Rosen defines them as systems
"in the ambience or external world".
 
Natural systems, in his parlance, are "actual, real systems" as opposed to models or simulations. They can be systems made by humans (like a house) or systems which spontaneously self-organize (like a hurricane). But he takes pains to point out that this is a very subjective activity: he says that science always is a subjective activity. Objectivity is a myth, at least the way we define it in science. Worse; the more we try to achieve it, the less likely we are to actually succeed. In any case, his use of the word "natural" in this sense refers to "real".
 
CL wrote:  I can't
understand Rosen's point very well, so I'd pleased to
know in what exact sense" the entailment structure of
the language itself can change by virtue of what the
language is about" and if Rosen effectively is
asserting this.
 
The point he is trying to make is that language doesn't exist in a vacuum: it interacts with that which it refers to-- and furthermore; that which it refers to has an impact on the language as well. This kind of interactivity relation between a system and its environment is exactly what drives evolution, for example. Languages evolve and organisms evolve. They do so in concert with environmental impacts and they impact the environment, in turn, as well. It never is a one-way thing. So the entailment relations, once again, commute: He was using language as a model of this aspect of biological system entailment relations.
 
Are the logical constants of the
formal language of a formal logic a candidate for
"entailment structure"?
 
I would say yes, if by "logical constants" you are talking about the rules of the formal language. For example, in mathematics a plus sign is the opposite of a minus sign, in operational terms (not talking about positive and negative numbers, here). Addition (the plus sign) is the operation for the process of putting things together and subtraction (the minus sign), conversely, is separating things out. This logic is set and always holds true within the language, so a plus sign isn't sometimes about addition and sometimes about division, say. It's always the same operational process. The internal rules specifying how the system works-- that's entailment structure. 
 
But most internal rules have referents to external "things". Even simple systems, like a car, carry semantic entailment information that isn't syntactical. For instance, cars have tires. They don't have pontoons or skis, they don't have a hull, like a boat-- they have tires. Tires refer to roads and gravity, among other things. Similarly, complex systems also have a great deal of semantic information about their environment encoded into them and they interact with their environments in ways that can radically change the environment in turn. Plants changed the composition of Earth's atmosphere-- because of the entailment structure of their physiology and metabolism. Beavers have webbed feet and a rudder-like tail, yet they're land mammals. Beavers are a really good example of systems which change their environment purely by following their internal entailment structure.
 
Well, it's getting late and I'm not sure if I'm still being lucid, so I better quit now and pick it up again tomorrow. I hope this has helped at least a bit?
 
Cheers,
Judith
 
BioTheory: An E-Journal of General Science in the Rosennean Complexity Paradigm http://www.rosen-enterprises.com/RobertRosen/BioTheoryLaunch.htm
Website address: http://www.rosen-enterprises.com/