Game Development Reference
In-Depth Information
Four Out of Five Pirates Surveyed…
In the Pirate Game, we questioned pirate A's decision because it doesn't fit with our
gut feeling of what he ought to do. Part of the reason for this is that our view of the
other
pirates may be a little off. One of the initial guiding premises of the set up of
the game was that “all the pirates are rational.� Once again, it is that
unreasonable
assumption of superrationality that disqualifies a purely normative approach from
believability. If the other pirates are
not
rational, then all of the calculations we did
above are meaningless. If they are not purely rational, then those messy things like
fear, spite, shame, and even simple calculation errors will creep into their decision
making. If their decision making is compromised and unreliable, then we can't accu-
rately put ourselves in their shoes, can we?
Therefore, whether or not we are purely rational is not the only question that
needs to be addressed if we are to formulate our approach. If we are rational and so
are the rest of them, all is well, but if even one of them is not rational, it can signif-
icantly skew any sort of predictive result.
Of course, in the Prisoner's Dilemma, it was in our best interests to work together.
In the Pirate Game, the competitors all want to launch each other overboard and
get more gold for themselves. Can we really assume they are both perfectly rational
and have
our
best interests in mind as well?
So, perhaps our quandary lies in that we are trying to solve the wrong problem.
Rather than trying to ascertain the “right answer� that optimizes our solution in the
perfectly rational world, we need to try to model behavior that takes into account that
people aren't purely rational with all of the information and calculation ability.
The answer to the question is more of a function of what
game problem
we are
trying to solve. Also, the answer may be affected by what it is we are trying to ac-
complish in this particular calculation. If we need mathematical accuracy, then the
normative approach is the best one to use. If we need psychological believability,
the descriptive one is preferable. And someplace in the middle we can draw traits
from both. In fact, only by juxtaposing the two approaches were we led to the fact
that the descriptive approach (e.g., being altruistic) was downright incorrect from
an efficiency standpoint. Only by analyzing the two of them simultaneously were
we able to determine that the normative approach (e.g., mathematical exhaustion)
didn't take into account potential psychological factors. That is, the “shoulds�
seemed perfectly viable until we brought in the data that suggested that real people
simply don't do what they “should do.�
So, maybe our gut instinct wasn't too far off. All of this wonderful math and
logic that computers speak so well and artificial intelligence (AI) programmers are
so fond of goes right out the window. If our pirates are to look and act
realistically
they may not act completely rationally. If they don't act completely rationally, they
