Game Development Reference
In-Depth Information
In the first group, most of the people chose A—that is, to save one-third of the
people (200) for
certain
rather than roll the proverbial dice and gamble about sav-
ing everyone yet risk losing
everyone
in the process. On the other hand, the second
group chose program D—apparently thinking that the odds of saving everyone
were preferable to heartlessly letting two-thirds of them (400) die. The contradic-
tion is clearly irrational, given that there is no mathematical or statistical difference
between the two courses of action: A/C vs. B/D (Figure 9.2). The only difference
was in the
perception
of the situation based on the carefully chosen language of the
questions. The words we used played upon people's feelings—the horror of losing
400 people when they could have been saved.
FIGURE 9.2
In Kahneman and Tversky's outbreak problem, people overwhelmingly
selected different choices based entirely on the wording of the question despite the fact
that the meaning of the words was exactly the same. (The selected answers are circled.)
If flawed (or at least skewed) judgments like these were used in constructing
mathematical ratings for Bentham's hedonic calculus, the results of any given
calculation would be all over the map. The premise that he had begun with, that his
method would cut through all the complexity of making multivariate decisions, was
hamstrung by the very fact that most of the variables are next to impossible to
define.
Put another way, it isn't the solving of the equation that is the problem; it is the
determination of what values to use when we are solving the equation that is the
most important part. When approaching the task of constructing behaviors, many
people stop at the point of coming up with an algorithm without considering the
fact that the numbers they are feeding into the algorithm may have been subjectively
skewed—even slightly—in such a way as to not yield the outcome they would like.
