Information Technology Reference
In-Depth Information
4.3 User Behavior and Free Rider Issue
Previous experiment results indicate the two methods (TF-IDF and CTD) could
provide satisfactory outcomes than traditional TF method. However, the differ-
ence between TF-IDF and CTD is not significant. In order to analysis the reason,
we go further to analysis the user behaviors when using i-Bike service.
According to the contribution rate in user's behavior log file, we found the user
behaviors could be unfolded into three types: active users, passive users, and free
riders. Active users are who will contribute information and provide feedback in-
formation energetically; there are about 13% of active users in the service system.
Passive users will still contribute information and feedbacks; however, the quan-
tity of information is comparatively less than active users. In this study, there are
27% of passive users in the service system. We found there is a significant free
rider problem in the service system. There are 40% users will contribute to the
system positively, but other user's behavior are shown as free riders. These free
riders do consider the provided proximity social intelligence for improving the de-
cision quality. However, they did not contribute or provide feedback information
to the service system.
Since i-Bike service is a Web 2.0 service that addressing on the concept of user
generated content, the less contribution may imply user's attitude to collaborative
behavior. The proximal social intelligence do requires the contribution of all pos-
sible information sources to enable the collative wisdom. However, the less feed-
back contribution and less information sources could diminish the power and in-
fluence the impact of social intelligence. In order to understand the difference of
the recommending mechanism, we evaluate the difference by exclude the free rid-
ers to determine the situations on those positive users.
After separate the free riders, analysis result indicate the TF-IDF method has
better satisfaction than CTD method. There are difference between two groups
was confirmed by independent t-test (p<0.1) in perceived usefulness dimension
and information quality satisfaction dimension. We can say that users who posi-
tively contribute to the service system could experience varied difference of the
two recommending mechanisms.
Although we found there exist some difference between TF-IDF and CTD rec-
ommendatory methods, but the variation is little. The original CTD method is pro-
posed to improve the TF-IDF method, but the effect is not clear in our case. It
could be influenced by the classification of the bicycle leisure entertainment that
reduces the effect in this particular field domain.
4.4 Managerial Implications
The goal of this chapter is to explore the proximal social intelligence for collabo-
rative recommendation. The design applied the web 2.0 service design that en-
ables users participate in the collaboration process and empower the experience
co-creation. Different recommendatory mechanisms were used for generating
suitable tags to enrich information quality to improve the decision quality. Ac-
cording to the experiment result, both TF-IDF and CTD methods could provide
