Game Development Reference
In-Depth Information
Great Sugarless Debate. We used a random number process in conjunction with a
carefully constructed probability distribution (i.e., four out of five).
Yes, the dentist example was simple, but we can do far more if necessary. In fact,
we explored some of this in Chapter 11 when we randomly created a population of
guesses in the Guess Two-Thirds the Average Game that was similar to the results
of the Danish study. Our approach in that exercise was to use random generation
of guesses based on a combination of specifically tailored probability curves.
Thinking further back to Chapter 6, recall the real lesson of the Guess Two-
Thirds the Average Game. We were trying to find the sweet spot between the purely
rational (yet very unlikely) answer of 0 and a completely random (and also unlikely)
guess. That is, we were trying to get away from the rigid rationality of normative
decision theory and move toward the more human-like descriptive decision theory.
Our solution at the time was the carefully crafted and controlled application of
structured randomness.
The question that we still must answer, however, is how do we build that struc-
ture around our randomness? In the Guess Two-Thirds the Average Game, we had
an example from which to work. It was much like how I was able to paint my some-
what realistic-looking pig because I was looking at a photo of a pig. From an artis-
tic standpoint, I couldn't draw a pig from scratch. However, by looking at one, I
could copy the bits and pieces and replicate what was already there. We did this in
Chapter 11 by reconstructing the guesser distribution one part at a time. We
weren't modeling
why
they were guessing the way they did, only that they did so
because a survey told us, “This is how 19,000 Danes guessed.�
Unfortunately, we don't always have a handy survey to tell us how people di-
vided themselves among the choices that range between “the best� and “the most
ridiculous ever.� We need to have a more procedural way of outlining the frame-
work that will support our otherwise stochastic process. Surprisingly, we are already
closer to the solution than we might think.
Let's break our problem down into the features we need in our decision model and
the features we don't want.
We want more variety than “the best� answer.
We
don't
want to include “the worst� answer.
We
do
want to include “reasonable� answers.
We don't want to include “unreasonable� answers.
