If
a consumer needs to choose among several options,
and he/she is not certain what to do, he/she will
often rely on word-of-mouth opinions. However, not
all opinions may be equally desirable. If the consumer
is, for example, a 60-year-old male, Ganesh, and he
is not certain whether he would enjoy a particular
book, he likely would not give much weight to the
opinion of his 12-year-old granddaughter. Similarly,
Ganesh may not give much weight to his daughter-in-law's
opinion either she being 25 years younger than Ganesh.
Still, he might give more weight to his daughter-in-law's
opinion than to his granddaughter's opinion.
Now
consider a situation in which Ganesh has Web access
to the opinion of thousands, perhaps, millions of
people concerning the book of (potential) interest
to him. That is, Ganesh, perhaps due to membership
in a Web-based book group, or by making a purchase,
or simply by considering the making of a purchase,
is able to make use of these opinions, at least in
some summary forms. One option might be to present
Ganesh with the average opinion of all the people.
Simply to be concrete, let's say this average comes
from a 1-5 scale, where 5 means "must read"
and 1 means "awful;" (this is the scale
used by the website, MovieLens, except, of
course, that "must read" is "must see").
However, the group of all people likely includes a
few people like his granddaughter, and many people
like his daughter-in-law. Wouldn't it be useful to
have some type of `intelligent agent' that could provide
Ganesh with the opinions of only those people who
have rated books Ganesh has previously read and, of
these people, only those whose ratings closely matched
Ganesh's own opinion of these books. Of course! Can
a system such as this actually be implemented? Yes!
It already exists in many settings. It is most often
referred to as collaborative filtering. |