Game Theory and the Golden Rule

jbone at place.org jbone at place.org
Tue Dec 2 06:36:37 PST 2003


JR and Eugene opine:

> > Computational parity between intelligent agents is one of those 
> pervasive
> > assumptions in human thinking and institutions based upon historical
> > reality, one that more or less asserts that 1) only humans are 
> intelligent
> > agents, and that 2) all humans have functionally equivalent 
> intelligence
> > (although this illusion frays at the extremes).
>
> It's very obvious, and truly remarkable how many people don't get
> that the game completely changes when the old rules no longer apply.

Let me qualify and expand on my statement a bit...  first, the 
reasoning I eluded to is helped by viewing the game series as infinite; 
  once you look at the series that way, there are only two 
possibilities:  playing the series as net zero-sum (in which case net 
payoff is zero) and playing the series as net positive-sum (in which 
case the net payoff is infinite.)  Differences in effective 
intelligence show up as mismatches in the rate at which the game can be 
played.  Assuming a willingness to play the game, this does *not* 
actually effect the payoff matrix at all.

Now, one might ask why the more intelligent being might want to play 
the game in the first place.  And that's a valid question, one for 
which we can probably have no good answer.  But as long as the being 
wants to play the game and views the series as potentially infinite, 
then it seems that playing for net positive is the preferable strategy.

So the qualification here is that, I suppose, there are really three 
possibilities:  either the being in question is going to treat the 
series of interactions as potentially infinite, with net positive 
cooperative strategy;  or they're going to view the game as finite, in 
which case we get a Terminator-like scenario (and, well, too bad --- 
probably unavoidable under some assumptions);  or (worst) the being is 
capricious and plays for perverse amusement in iteratively winning 
net-zero interactions.  Given that either of the latter two are 
probably unavoidable under various assumptions in any interaction 
between vastly different intelligences, I see no point in losing sleep 
over them;  OTOH, it seems to me that most intelligent beings will see 
that even the occasional net-positive interaction is a boon, and will 
treat the series and infinite and play for net-positive.

jb



More information about the FoRK mailing list