Game Development Reference
In-Depth Information
Numbers such as the “four out of five� statistic not only tell us about the behav-
ior of populations, but they offer us useful tools with which we can reconstruct the
behavior of populations. If we were to construct a histogram about the dentists the
same way we did with the number guessers, it would look like the one in Figure 11.1.
FIGURE 11.1 A simple histogram showing how the marketing adage
“four out of five dentists surveyed� would look.
Based on this histogram, if we were to ask what was the percentage chance of
randomly selecting a dentist who answered “yes,� the answer would be four out
of five. Therefore, if we were to build an agent pretending to be a dentist, and we
wanted this dentist to answer the “sugarless gum� question, we could use this exact
same histogram as a source. In this (admittedly simple) case, we would leverage that
same mathematical proportion—four out of five… or 80%.
Turning the Tables
The trick is in turning the statistics back upon themselves. By looking at the survey
data, we can realize that we are 80% likely to encounter a dentist who recommends
sugarless gum. Therefore, when we create a random dentist, we make him 80%
likely to recommend sugarless gum. We are taking survey data from five dentists
and boiling it down into collection of probabilities that we can use to construct one
dentist at a time.
To reiterate, in this case we created the histogram from the knowledge we had
of dentists (i.e., “four out of five�). By using that histogram as a decision tool for
our dental agent, we applied the process in reverse. We gave a four in five chance
that he would respond “yes� and a one in five chance that he would respond “no.�
If we change the numbers, the histogram would change. Of course, if we changed
 
Search WWH ::




Custom Search