Game Development Reference
In-Depth Information
So, if we were to put this into a game situation, which method do we use? The
normative one provided us with the answer of what a pirate
should
do. The descrip-
tive one provided us with the answer of what pirates (or at least pretend ones)
tend
to do. The former is the optimum solution; the later is the more “realistic.�
S
UPERRATIONALITY
In the Prisoner's Dilemma, we touched on the difference between the strictly dom-
inant strategy of defecting and the Pareto optimal strategy of keeping quiet. In that
case, the only way the optimal strategy could be accomplished was if our partner
elected to stay quiet as well. The only way we could feel comfortable making that
decision was if we knew that our partner would recognize the optimum strategy…
and believed that
we
were going to follow it as well. Therefore, the important issue
is not simply one of
us
behaving rationally but also acting under the assumption
that all
other
players act rationally as well. This is known as
superrationality
, as coined
by Douglas Hofstadter in one of the eclectic articles in his 1985 book,
Metamagical
Themas
(ISBN 0-465-04566-9).
Admittedly, making a rational decision whose success must be based on the fact
that everyone
else
in the room is also making a rational decision is sometimes a bit
of a stretch. In fact, in some cases all it would take is
one
person to not act rationally
to send the entire framework into a tumble. As we will explore later, this is one of
the greatest handicaps in normative decision theory as discussed in Chapter 4.
Remember that the requirements for normative decision theory are:
Has
all
of the relevant information available
Is able to
perceive
the information with the accuracy needed
Is able to perfectly
perform
all the calculations necessary to apply those facts
Is perfectly
rational
This is all well and good if we are dealing with a puzzle that does not involve
other thinking agents. However, merely by their inclusion, we are potentially dis-
qualifying the first two items on the list. Short of some form of omniscience, we
can't read the other players' minds—therefore putting all the relevant information
behind a screen of doubt. Even information we can perceive will not necessarily
be completely accurate. As we can see, while normative decision theory and the
“shoulds� that it spits out for us can be rather helpful in solving some sorts of prob-
lems, it begins to show weaknesses when other people are involved.
