Game Development Reference
In-Depth Information
to be defined. Eventually, we can arrive at one value that represents the total utility
that the agent would derive from the action. This concept is known as
multi-
attribute utility theory
(MAUT).
MAUT is used in many industries as a decision tool, although it may go by dif-
ferent names (e.g., multi-criteria decision analysis, multi-criteria decision making).
The business world presents an overflowing cornucopia of formalized methodolo-
gies available to assist decision-makers in processing massive amounts of data. A
quick search even turned up the
International Society on Multiple Criteria Decision
Making,
which claims to have over 1,400 members in 87 countries. On their Web
site (
www.terry.uga.edu/mcdm/)
, they define their field as follows:
“MCDM can be defined as the study of methods and procedures by which
concerns about multiple conficting (sic) criteria can be formally incorpo-
rated into the management planning process.�
Aside from the egregious spelling error, that definition pretty much sums up
what we are trying to accomplish with game AI. Working backward, “management
planning process� is a buzzword for making a decision. To arrive at that decision,
we need to “formally incorporate� a number of “conflicting criteria.� That sounds
strikingly like what Bentham had in mind.
This approach is also at the core of designing and programming game AI. If we
look back over the examples we used earlier in this topic, we find that MAUT would
have been appropriate in many of them. They use many, possibly conflicting crite-
ria that we need to formally incorporate into a unifying algorithm to arrive at a
decision. For example:
When my daughter was formulating her strategy for running for VP of the
fifth grade, she identified many criteria that would have led to decisions about
whether to run for president or vice president, whether or not to make stickers
and posters, and if she should tailor her speech to be more kid-friendly.
When my son was considering raising prices in his zoo, he pondered the per-
ceived value of his zoo based on how many exhibits he had already created and
the balance between more visitors at a lower price and fewer visitors at a higher
price.
When I analyzed my decision to use the next razor blade in the pack, I took into
account how many blades were left, the cost of purchasing more blades, my
comfort in shaving with a used blade, and how much I had already used prior
blades.
