From [Marc Abrams (2005.02.08.1300)]
Something has been bugging me for a very long period of time and it has only been recently that I have had the opportunity to understand and address the issues involved.
I have been involved in SD modeling for over 20 years. I am not an expert modeler, but I can throw a decent model together. But something always bothered me about the consistency of the effectiveness of the models. Some models ‘worked’ well and others weren’t worth a crap, and I’m talking about work done by the same modeler. By ‘work’, I mean the models were effective in representing what they purported to represent. That is, they were effective predictors of the processes they modeled. I could not understand what the problem was.
Now I do, and it resides here on CSGnet, BIG TIME, and it has to do with the reasoning and logic involved in modeling which of course is mathematical, or deductive.
Descartes has been responsible for some wonderful (Cartesian plane), and some woeful (dualism) ideas. Another idea of his that was adopted as a standard well into the 20th century, and still very much alive is the notion of mathematical logic. That is, deductive reasoning, or formal reasoning. This has ruled ‘scientific’ thought for over 300 years, and it has been a huge mistake. Oh, not the logic itself. That is wonderful. Just the idea that it could be generalized to all types and kinds of reasoning is where it collapses, even with regard to science.
Let me try and explain why;
From here on in I will use the term ‘deductive reasoning’ to be synonymous with either ‘formal logic’ and ‘mathematical reasoning’ , and I will use the term ‘inferential reasoning’ to mean ‘informal logic’ or ‘inductive reasoning’
Mathematical modeling is predicated on the notion of deductive reasoning and the use of syllogisms.
That is;
A. The conclusion follows necessarily from the premises
B. The conclusion contains no information not already present (at least implicitly) in the premises.
C. These properties suggest two corollaries.
-
Deductive reasoning [modeling] is analytic; it requires no reference to the external world and it may be counter-factual.
-
Deductive reasoning [modeling] does not add to our store of knowledge; it merely rearranges it.
The major syllogisms;
A. Categorical syllogisms contain statements that relate categories to other categories.
B. Conditional syllogisms begin with an “if-then” statement.
C. Disjunctive syllogisms begin with an “either-or” statement.
Formal logic also deals in CERTAINTY. The basic mathematical axioms are all considered to be ‘facts’, and a mathematical proof is nothing more than working backward toward the base of accepted axioms (truth).
Unfortunately for us, the world, and human behavior does NOT have the properties of mathematical axioms. We live in an uncertain, probabilistic world. Round pegs often don’t fit into square holes.
Now, until fairly recently, formal reasoning was regarded as the prototype for all forms of reasoning. It was seen as the model case, and what other reasoning ought to do is it ought to aspire to the standards of formal reasoning. This view has come under challenge, and it’s come under increasingly strong challenge, and the challenges take several forms. One of them is to suggest that very seldom, does one actually reason in syllogistic form.
Remember I said it requires no reference to reality. It could be done entirely in symbols, and often is, All A is B. All B is C, therefore all A is C.
How often is that the case, when we talk about things, when we make claims, that we can abstract from reality and it makes no difference what we are talking about? Very seldom. In fact, most of the time_precisely_ what we want to talk about is the subject matter categories we are discussing.
Consider another feature of syllogistic reasoning, of formal logic. Remember I said if we want to measure, our categories are; all, some, and none, and those categories admit no degrees. It makes no difference whether it’s one percent, 99 percent, or anything in between. In studying human behavior how often is this the case???
This my fiends, in a nut shell is one huge problem. It also represents one of the major problems with modeling.
The numbers we use in models represent what??? That is, what bit of certainty do these numbers purport to represent.
Another and related criticism is to suggest that most reasoning, is not represented well by a form in which the conclusion contains no new information. What do we want to do most of the time when we reason? We want precisely to get from something we already know to something we don’t. We want to get some new information. We want to move from premises that we’re fairly confident of and see if they lead us to conclusions that’s new that we can also be fairly confident of. And this movement-- from what we know to what we don’t-- involves a leap of faith, which an arguer tries to justify with evidence. That my friends is called DATA.
A model is absolutely USELESS without the data to back it up because as I have shown, it may, or may not represent anything in the real world.
Rick’s models are a fine study in logic, but are MEANINGLESS as tools to show that the actual properties and processes the model purports to represent actually exist. We need DATA for that and Rick has ZERO, ZILCH, NADA, for his spreadsheet model. What he has is a LOGICALLY accurate model of SOME system. Whehter or not it repesents a human control system of some type is quite a LEAP IN FAITH. But as I said yesterday. Faith is not in short supply with Rick or his models and I am happy for him because fior his models he really needs hiis faith and will continue to need it until he comes up with some DATA.
But this is going to be a REAL problem for Rick. because Rick has NO idea of what kind of data he NEEDS for his spreadsheet model. NO clue in the world and he will NEVER, EVER find any in his DEDUCTIVE_MODELING.
Inferential reasoning leads to discovery, NOT deductive reasoning.
If you are interested in doing proofs I suggest you stick with mathematics, if you are interested in discovery I would strongly suggest some understanding of inferential reasoning.
George Polya, the wonderful mathematician has a wonderful 2 book set on it, Mathematics and Plausible Reasoning. I also strongly believe that the concept and technique of ‘argumentation’ provides a structure and method for acquiring the requisite data necessary for model building.
I will be implementing it on my Yahoo site.
All are welcome to come join the adventure. Contact me for details if interested.
The goal is a validated behavioral control model.
Marc