Re: A whiff of reality...

[Gary Polhill]

"glen e. p. ropella" wrote:
> 
> At 06:26 PM 5/2/00 +0100, you wrote:
> >
> >Assume: If any model is valid, then it will predict the data
> >Assume: My model predicts the data
> >Conclude: My model is valid
> 
> This seems pretty odd to me.  Validation consists of:
> 1. taking data off of two systems
> 2. comparing and contrasting the two data sets
> 3. quantifying the degree to which the two data sets are different
> 
> The "my model is valid" decision lies inextricably in the
> observables and the tolerances for sameness and difference.

Presumably the logical fallacy above assumes a perfect prediction (I don't
want to speak for Oreskes et al, but let's say that's what they mean) -- 
this is just zero tolerance for difference. Surely the fallacy of affirming
the consequent still applies, however, even if you bring in the fact that
the decision to call a model valid is made on the basis that the observations
of the two systems falls within your tolerance for calling them the same.

1. For all valid models, the data taken from the model and the data
   taken from reality are within the observer's tolerance for saying they
   are the same.
2. The data from my model and from reality are within my tolerance for
   calling them the same.

Therefore: My model is valid.

I think maybe the problem is that it is always possible to have a model
that predicts the observables, but is not a "true" representation of
what is going on in reality... A bit like the Ptolemy/Copernicus models
of the solar system Cal mentioned earlier.

Or maybe the real problem is not imperfect models but imperfect
observations....?

> 
> It's immanently clear that many people don't understand two
> things about validation and verification: 1) that they are
> inextricably tied up in the observables (Smith, Popper, Maturana, etc.)
> and 2) that they differ only in the subjective ontological status
> of their referents (i.e. system A is "reality" whereas system B is
> "a computer program... the only difference between "reality" and a
> "computer program" lies in the scientists mind.)

Not sure about point 2. Isn't the whole point of building a model
to give you some kind of predictive/explanatory power in reality?
After all, you are not trying to establish the validity of reality
against your model (unless, perhaps, you are an economist ;-) ...).
OK OK, so maybe from the point of view of validation there is no
difference -- you've just got two sets of data, but (and call me
an egoist), I think the difference between their origins (though it
is only in my mind) is important.

Gary

-- 

Macaulay Land Use Research Institute, Craigiebuckler, Aberdeen. AB15 8QH
Tel: +44 (0) 1224 318611               Email: address@hidden


                  ==================================
   Swarm-Modelling is for discussion of Simulation and Modelling techniques
   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
   please send a message to <address@hidden> with "help" in the
   body of the message.
                  ==================================