Friam Digest, Vol 38, Issue 3

Posted by Marcus G. Daniels-3 on Aug 06, 2006; 11:12pm
URL: http://friam.383.s1.nabble.com/Friam-Digest-Vol-38-Issue-3-tp522318p522340.html

Phil Henshaw wrote:
> I think modeling is out of reach, but story telling may not be.  Telling
> the stories of how complex events can be read or misread would be a real
> service.
There will be policy makers and I think it is safe to say they'll find
it easier to convince people of their policies if there are some
dramatic stories involved (e.g. 9/11, WMDs).  I expect a careful and
restrained story of the kind you describe above will be overwhelmed in
general by story tellers at think tanks like the Project for the New
American Century who don't hesitate to provide `leadership' (Perle,
Wolfowitz, Rumsfeld).

On a technical note, I don't buy that social simulations would be
computationally prohibitive, given the will.  The fastest general
purpose supercomputer at Livermore is $100e6 U.S. (BlueGene/L) having
130k processors.   Suppose a simulation ran for a day, that's still 130k
simulations a day.   That's a lot of sensitivity analysis one could do.  
It might take 10 teams of modelers to keep such a machine busy.   For
national security, what's a $100 million here or there?

The 2006 budget for Advanced Simulation and Computing Initiative
computing was $661 million and $6.3 billion overall for stockpile
stewardship.  Yet I keep hearing that `non-state actors' the new threat..
> How do you model brains full of made up nonsense??
Detectives, trial lawyers, and spies tease out models from deceptive
people and suboptimal evidence.   No shame in formalizing these models,
if only to make it clear what is far from being known.  And to deal with
a culture that only wants compliance and to stay `on message'  all I can
suggest is to 1) stomach it, and 2) slowly bend the message in some
other direction.

Marcus