Game Development Reference
In-Depth Information
11
Probability Distributions
As we discovered in earlier chapters, there are many reasons for the gulf
between the one “should be done� solution that normative decision theory
provides and the myriad possibilities that are summarized by descriptive
decision theory—that is, what people as a whole tend to do. The suggestions and
guides that descriptive decision theory provides often are based on observation
and collection of information. This data is necessarily an aggregate. As such, we have
a picture of the population as a whole rather than any individual member of that
population.
The end result is that, while we know that people do different things, we don't
know
why
they do them. While we can ascertain to some degree some of the
possible
non-optimal
solutions that people may provide to a particular decision,
often the reasons
why
they do so get lost in the background noise of human indi-
viduality.
The other shortcoming of the descriptive decision theory approach is that, for
obvious reasons, the game development world doesn't have data on the behaviors
we would like to model. We simply don't have enough data on how often orcs elect
to use their hand axes or the accuracy with which demons from hell can toss balls
of flaming plasma. Even in cases where we
do
have data from which to work, it may
not provide the sort of behavior mix that makes for a believable, engaging
game
character. Therefore, to model these behaviors, we must be able to assemble our
own substitutes for the data that we believe is representative of the population as a
whole.
Throughout this section, the term
population
may represent a group of data
points rather than simply the typical definition of “a group of people.� In large
part, these concepts can be interchanged. For example,
population
could signify
many people each making one decision about
X
or
one
person making
many
successive decisions about
X
.
241
