[Osx-nutters] Denial

[David Cake]

At 1:36 AM +0100 17/8/07, Stefano Mori wrote:
>On 2007-Aug-17, at 01:26, David Cake wrote:
>
>>>  Then what do you think is a good method to predict the future of a
>>>  complex system?
>>
>>	I think he was saying that its condescending to assume the
>>  rest of us just underestimate it.
>>
>>	 The term complex systems in the scientific sense (I'm even
>>  published in the proceedings of a Complex Systems conference, back in
>>  1992) means a system that is between rigidly predictable, and pure
>>  random stochastic systems - ones whose behaviour is predictable in
>>  some ways, but not others. Strange attractors, chaotic behaviour, all
>>  that sort of stuff.
>>	And we, of course predict some aspects of the long term
>>  behaviour of complex systems regularly. The weather is a complex
>>  system, and so its impossible for me to predict whether it will rain
>>  on this day next year. I can give you a pretty good probability that
>>  it will, based on past behaviour, though.
>
>
>So some aspects of the behavior of more complex systems may be 
>predictable in some ways but not others.
>
>What's a good method for working out which behaviors are predictable,
>and which methods are good methods for predicting those behaviors?

	I've been trying to think of a good summary answer to this, 
and I've pretty much failed to come up with a better answer than 
"maths".


	I know that sounds dismissive. Its not intended to be.

	"Maths" remains a good solid answer - but I think part of the 
issue here is that you need a pretty large amount of maths to get an 
intuitive feel for what mathematical analysis of models can and can't 
do, and even more maths to be able to actually explain how it might 
best be done, and I fall somewhere in between these two points (I am 
confident would have reached the latter by now had I continued 
following the academic path I was on then, but I didn't).
	One mathematical test that is relevant (but not the whole 
story - stats obviously has a big part to play in models partially 
based on stats data, and my stats knowledge is minimal) is the 
Lyapunov number, which is a mathematical test for chaotic behaviour. 
That particular bit of information leapt out of my memory, which was 
a surprise considering I last looked at this field more than a decade 
ago, I know there are other tests, but they failed to likewise leap.
http://mathworld.wolfram.com/LyapunovCharacteristicNumber.html
	Now, I'm trying to continue the debate here, or throw in 
random mathematical facts and pretend its an argument. I think I'm 
rather trying to demonstrate, that while intuitive qualitative 
argument about complex mathematical models might be occasionally 
entertaining, the real debate about their validity will be a 
mathematical one, which none of us are really equipped to have (the 
few of us who might have the knowledge almost certainly don't have 
the time).
	And ultimately, this is why I think the political or 
philosophical questioning of science is often pointless - often those 
asking the questions are poorly equipped to understand the answer.
	Regards
		David