Game Development Reference
In-Depth Information
sports teams over the course of many years. Just because a team has a history of
success over the years, that doesn't mean they are going to do that well during the
upcoming season. (Or, in the case of my Cubbies… uh… never mind.)
Another problem that can occur is when we end up in a cyclical arrangement.
That is, when the availability of data on what has been done by others in the past is
used to drive the decisions of others in the future. At times, this seems like it would
be a good idea. If I were a dentist trying to decide on what sort of gum I would
recommend to my patients, I may look at what
other
dentists before me have rec-
ommended. In our clichéd “four out of five� scenario I may decide that sugarless
gum must have merit enough that my colleagues are suggesting it to their patients.
Therefore, I should do so as well. There are problems with this approach, however.
What if They Are Wrong?
As we covered above, we have to assume that the data is relevant and well informed.
Sometimes it is not. There is a well-known (and oft-repeated) phenomenon in pol-
itics, wherein polling data is used as news. If we were to see a poll on an issue or a
person who told me that a majority of people felt a certain way, I might be inclined
(like the dentist above) to assume that there is something to the position. As time
passes and more people believe the data and shift their opinions accordingly, the
poll itself shifts more. This may not be due to any sort of legitimacy of the topic in
question. It may be entirely due to the fact that people trust each other's opinions
and think that the next person (as reported by the poll) must know more about the
issue than they do.
The more people that are reported as supporting the position, the more
strongly my conviction that there must be some relevance or benefit to it, despite
the fact that I may know nothing about the issue. If we were to make a decision
based on that information, we would have committed a grievous error. We have
taken descriptive decision theory (the polling data) and used it as a replacement for
the prescriptive decision theory (what we
should
do with the information).
As I related in Chapter 3, my daughter Kathy noticed the effect that the descrip-
tive data had in the fifth-grade election. People were willing to base their support
for a candidate on the knowledge that a group of friends supported that candidate
as well (i.e., the survey). For what it's worth, in the milieu of a fifth-grade student
council election, there
is no
prescriptive decision theory that can be presented.
There is no reason that you
should
vote for any one kid over another. In that case,
the perceived opinion of the group is good enough. In Kathy's observation, each
kid believed that the other had a perfectly valid reason for supporting that candi-
date… and they never bothered to compare notes. It was an exhibition of the social
momentum of peer pressure at its finest. You wouldn't need Hari Seldon's psycho-
history to predict the voting habits of 10-year-olds.
