Trust in Online Collaborative IS - Community-Built Databases: Research and Development - page 366

14.6 Discussion and Conclusion
In this chapter, we summarized several studies of Wikipedia as the largest online
collaborative information system in order to a analyze the issue of trust. We showed
how user reputation can be modeled according to his/her edit behavior pattern. We
also showed how this user reputation value can be used to assess collaborative
content quality. One application of this work can be in scientific mashups, espe-
cially when the content comes from collaborative information repositories
where the quality of such content is unknown. In [ 58 ], we studied CalSWIM as a
watershed scientific mashup whose content is taken from both highly reliable
sources and Wikipedia, which may be less so. We showed that how integrating
CalSWIM the reputation management system can help to assess the reputation of
users and the trustworthiness of the content. Using user reputations, the system
selects the most recent and trustworthy revision of the wiki article rather than
merely the most recent revision, which might be vandalistic or of poor quality.
Although this study of the problem of trust in the CalSWIM mashup indicates
that the idea of showing the most trustworthy and recent revision of an article to a
user can be beneficial for fetching content from wikis. However, it is important to
note that assessing trustworthiness of content based only on the reputation of the
contributor has some limitations:
l Data sparsity: for a considerable number of users of Wikipedia, we do not have
enough information for accurate reputation estimation. The models that we use
for estimation of user reputation are based on the observed behavior of users and
how other users react to the contributions of these users. Therefore, in cases
where a user is new to the system, we do not have a stable reputation estimate
for him.
l Anonymity: a significant number of users contribute to Wikipedia articles
anonymously and they are identified only by their IP addresses. However,
there is a loose correspondence between the IP addresses and the real-word
users.
l Expertise: quality of the contribution of a user to a topic depends on the expertise
of the user on that topic. Having one reputation value may not be a perfect
representative for quality of the contributions of the user on different topics. In
CalSWIM, we tried to alleviate this problem by estimating the reputation of
users based only on their contributions to water-related articles.
In addition to the above limitations, there is no guarantee that users will not
change their behavior in the future. So, a user who has contributed high-quality
content in the past might contribute low-quality content in the future. In addition,
when a new user comes to the article and contributes high-quality content, the
system sacrifices freshness for trustworthiness, only because it does not have an
accurate estimate of the user's reputation. This problem becomes worse for articles
that are updated less frequently. In the case of our CalSWIM mashup, some articles
are updated very infrequently. The average time span between submission of the