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

Re: Game Theory and its Relevance to Cypherpunks



At 21:57 20/07/96 -0700, Timothy C. May wrote:
>
>Fortunately, the goal of MAD was to
>not have to be used, and it appears now to have worked quite well (albeit
>at high cost).

My problem with applying game theory to complex situations like the 
competition between powerful nations is that it is too simplistic. IANAE, 
of course, though I have done some control theory, and know how 
complex the modelling of any system becomes if it contains non-linearities, 
delays, etc. In a closed-loop system, i.e. with feedback, trying to predict 
behaviour without the foggiest notion of how to quantify the impact of
Kennedy's grandstanding on the Kruschev mind (for instance) is questionable.

To suggest that MAD worked well on the basis of the limited tryout we gave it
has little validity. If it hadn't, we wouldn't be here, would we?
We are trying to
draw general conclusions based on a biased sample of one.

Reminds me of this committee of the British Royal Air Force, trying to find
ways to protect its planes better against German anti-aircraft guns. Someone
proposed putting an extra layer of armour on those areas that received the most
shelling, and most people seemed to like the idea. One upstart suggested the
exact opposite: putting extra armour on those areas which had received the least
shelling. "Remember," he said, "we can only examine the planes that came back."

I also have a problem with the cost you mention. What is
"winning" in the context
of nations? The arms race wiped out the Soviet Union, and arguably seriously
hampered the competitiveness of American industry (which was No.1 at the end of
the war), allowing countries like Japan and Germany, with far
smaller defence budgets,
to overtake industrially.

>Useful sources:

Thank you for the tips: I will check them out. Game theory is
fascinating -- I'm just 
not sure how applicable it is in formulating policy. Didn't work
too well in Vietnam...

Arun Mehta Phone +91-11-6841172, 6849103 [email protected]
http://www.cerfnet.com/~amehta/  finger [email protected] for public key