From: Anders Sandberg (asa@nada.kth.se)
Date: Fri Nov 22 2002 - 03:21:03 MST
On Thu, Nov 21, 2002 at 04:08:37PM +1100, Brett Paatsch wrote:
>
> Yeah. Now we moving in on some contempory stuff. If attacks _are_ more
> tempting, isn't the stability of the game decreased (though perhaps not past
> the point of manageability which may be your point)?
>
> With "nuclear terrorists" or non total nuclear war scenarios we really have
> a different game. One that is more complex though less risky in global
> terms, I think.
The "nuclear terrorist game" matrix looks like:
Cooperate Defect
Cooperate 1,1 -X,1
Defect 100,-100 100-X ,-100
The terrorist player is horizontal. If both play nice, they get one
year of peace. If the terrorist attacks while the nation player is
nice, the terrorists get their fireworks. If the nation player
defects by doing something the terrorists hate (invades nation X)
the terrorist will be upset by X. For the terrorist the rational
choice is to defect unless X is very large (and as has been remarked
before, certain belief systems might preclude this - the slaughter
of the faithful is surely just a part of the final tribulations and
just something to endure). For the nation player there is little
difference between cooperating or defecting, because the terrorist
makes his move independently.
This is a nasty game, because it makes defection very stable. You
can only "win" against rational terrorists where they have something
they desire more than destroying you.
Of course, real life is more subtle. There are other moves like
trying to find the terrorists beforehand, and the above game might
lead to a new game ("You killed my brother!" ;-) where revenge for
the previous actions now is part of the payoff (let Y be the amount
of satisfaction the nation player gets for taking revenge):
Cooperate Defect
Cooperate 1,1 -X,1+Y
Defect 100+X,-100 100,-100+Y
And so on ad infinitum...
Multiplayer games of MAD has some interesting effects - the players
who doesn't participate in a mutual nuking will benefit (or at least
not be harmed), and might get a safer world. So it would be to their
advantage to get others to blow each other up (shades of James Bond
villains).
On the other hand, many rational players will want to change the
payoff matrix to make disasters less likely. If Eurasia, Oceania and
Southeast Asia are rational but distrust each other, they might
agree to lock their missiles on each other so that if Eurasia fires,
it will nuke both the others regardless of who angered them (and
similarly for the others). That way one gets a stabile MAD for
three. Even if one "defects" by being a bleeding heart liberal and
only nuking their real enemy, they will be nuked anyway if they
fire.
I'm starting to get into the mindset of cold war planners. A rather
eerie experience.
> > Yes, but EFAE is too retaliatory. It produces endless,
> > non-constructive vendettas. TFT and Pavlov are more forgiving.
>
> EFAE _is_ to retaliatory. Is "Pavlov" a game named after Pavlov of
> "classical conditioning"/"Pavlov's dog" fame or are you refering to the
> phenomenon of "classical conditioning" itself?
It is the win-stay, lose-shift strategy: if whatever you did worked
well last time, do it again. If you lose, shift to the other move.
-- ----------------------------------------------------------------------- Anders Sandberg Towards Ascension! asa@nada.kth.se http://www.nada.kth.se/~asa/ GCS/M/S/O d++ -p+ c++++ !l u+ e++ m++ s+/+ n--- h+/* f+ g+ w++ t+ r+ !y
This archive was generated by hypermail 2.1.5 : Wed Jan 15 2003 - 17:58:18 MST