Trust in Online Collaborative IS - Community-Built Databases: Research and Development - page 343

understand the factors that lead people to freely share their time and knowledge
with others [ 17 , 18 ].
The positive correlation between content quality and user participation has been
discussed in some work [ 19 , 20 ]. Some studies also showed that building a good
reputation/trust can be a motivating factor that encourages user participation in
collaborative systems, as well as an incentive for good behavior [ 12 , 21 - 24 ]. There
is an extensive amount of research focused on building trust for online communities
through trusted third parties or intermediaries [ 25 , 26 ]. However, it is not applicable
to all online communities where users are equal in their roles and there are no
entities that can serve as trusted third parties or intermediaries. Reputation manage-
ment systems provide a way to build trust through social control without trusted
third parties [ 27 ].
A reputation management system is an approach to systematically evaluate
opinions of online community members on various issues (e.g., products and
events) and their opinions concerning the reputation of other community members
[ 28 ]. Reputation management systems try to quantify reputation based on metrics to
rate their users or products. In this way, users are able to judge other users or
products to decide on future transactions. A well-known example of reputation
management is eBay's auction and feedback mechanism. In this system, buyers and
sellers can rate each other after each transaction by using crude +1 or 1 values so

that the overall reputation of a trustee becomes the sum of these ratings over the last
6 months. Besides assigning these ratings, users can add textual annotations to
present their experiences during their transactions [ 29 ]. In other distributed envir-
onments such as peer-to-peer (P2P) file sharing networks or grid computing,
users can rate each other after each transaction (e.g., downloading a file). So far,
a considerable amount of research has been focused on the development of trust/
reputation models in virtual organizations, social networks, and P2P networks
[ 30 - 33 ].
Reputation management systems are difficult to scale when they have limited
sources of information. Users do not always give feedback about other users/
products. They also prefer not to return negative feedback [ 11 ]. To overcome this
problem, these systems consider reputation as a transitive property and try to
propagate it, in order to have an estimation of unknown users and products. In
this way, there is a high risk of propagating biased or inaccurate ratings. A study
of P2P e-commerce communities confirms this issue and shows that reputation
models based solely on feedback from other peers in the community are inaccu-
rate and ineffective [ 27 ]. To alleviate this problem, a reputation management
system can make its judgments based on objective observations rather than using
explicit experiences from other users, for example, by tracking behavior of users
in the system or analyzing users' feedback to products over time. Quite unlike
some research lines that are based on subjective observations in wiki systems
[ 12 , 34 ], in this work our aim is to quantify reputation based on objective observa-
tions of the users' actions.
In the open editing model of Wikipedia, users can contribute anonymously or
with untested credentials. As a consequence, the quality of Wikipedia articles has