Variation in Choice - Behavioral Mathematics for Game AI - page 422

Not only is rigid predictability boring to us, the observer, but it is not even re-
motely human. If we think of a situation in our own lives that occurs regularly, it is
likely that we can also make a list of our actions or reactions to that situation at dif-
ferent times. For the sake of continuity, let's look again at the list of bank robbery
reactions above. How many of them could we see ourselves possibly exhibiting?
Some of them might be more likely than others, of course, but we would have at
least a small chance of doing a great number of them. There also may be a few that
we may never consider. (I can no longer see myself jumping over a desk, for example.)
More importantly, if we found ourselves in a bank robbery situation again later
(hopefully not the same day!), would we pick the exact same reaction? We still
would have the same list of things we might do, but could we guarantee that any one
of them is the one we would do every time ? It is far more likely that we would react
different ways each time we were faced with a bank robbery situation. If we experi-
enced ten different robberies that were all similar in nature, how many different
reactions would we exhibit? Two or three? Five? Would we have ten different reac-
tions? (At this point in our saga, I am thinking that banking exclusively online
would also start to look more attractive.)
The quandary becomes apparent when we juxtapose having all those choices
that we might select from against a formula or algorithm that tells us that this is “the
best� reaction to have every time . Just as it was a false assumption to think that every
person in the bank would react the same as every other person one time , it is likewise
a tenuous assertion to say that one person would react the same way every time.
E MBRACING R ANDOMNESS
Individual people make decisions based on a massive collection of information—
only some of which is the current environment. The adage goes that “we are the sum
of our life experiences.� Those experiences were, themselves, information at one
time. Now, however, they have taken on a different role in our lives. What were
events at the time have become memories. Experiences have shaped beliefs. Traumas
have begotten fears. Pleasures have nurtured cravings and desires. Put into mathe-
matical terms, those life experiences of the past are now the formulas, equations, and
algorithms through which we pass the inputs of the current environment.
The problem is, even if we were to analyze our own life in great detail, we
would be hard-pressed to codify our own psychology accurately enough to
construct the algorithms that would yield the “right way� for us to respond to those
inputs. Even if we could do so, the construction of such a system would be prohib-
itively labor-intensive. (Yes, this is a massive understatement.) Additionally, the
computations involved would be processor-intensive. Oh yeah… don't forget that