Kollack [ 46 ] discusses the main properties that reputation management systems
should have to be effective tools for informing others of users' trustworthiness.
First, the cost of providing data to impact a trustee's reputation should be low,
hence allowing the collection of large volumes of data. Second, there should be
many alternative trustees, hence reducing the dependence on any single trustee for a
specific good or service. Third, the cost of impersonating a trustee and gaining good
reputation should be high. This means that good identity management systems are
needed for both trustors and trustees. Additional mechanisms are needed to make
sure malicious agents cannot alter a good seller's reputation. Finally, the reputation
method should itself be trustworthy, ensuring that it is unbiased and secure.
Reputation and trust management for reputation have been topics of intense study
in the computing literature [ 35 , 40 , 41 , 42 ] and in information sciences [ 6 , 17 , 18 , 19 ].
Reputation management methods are often designed for online e-commerce sites
that allow buyers and sellers to interact with each other. The main goal is to provide
buyers accurate information about the sellers. These systems rely on feedback
from buyers about the performance of the sellers, both positive and negative.
This information is then aggregated and presented to the buyers as an estimate
of the sellers' trustworthiness. Many design decisions impact the computation of
reputation: how often buyers are allowed to evaluate sellers, how information is
presented to buyers to seek their feedback, how information is aggregated, whether
past information can be forgotten or changed, whether the reputation management
system can plant fake ratings, and how reputation information is presented to the
buyers. Reputation systems may seek information about specific transactions and
aggregate these to compute reputation. Some reputation systems also allow users
to directly rate each other for trustworthiness and use this to compute reputations.
In this case, the trustworthiness of the rater can be taken into consideration when
considering her input.
In e-commerce sites like eBay, trustors are not typically concerned with assessing
trustees' ability; the transaction itself is usually not hard to accomplish. However,
the integrity of the trustee is in question. Do they represent the product that they
are selling truthfully? Will they ship the item quickly? Most reputation systems
designed in this context are based on sanctions. When a seller behaves badly, this
must have a serious impact in his reputation. In fact, a theoretical result shows that
aggregating the past history of a trustee does not improve the effectiveness of the
reputation system from an economic point of view [ 17 ]. This is similar to how
people evaluate the “trustworthiness” and “friendliness” of other people: negative
information is considered more diagnostic. It is irrelevant if the person behaved
well many times; if they betray trust once, they are not trustworthy.
A second type of reputation system is designed to judge the quality of the
service provided by a trustee. This is especially important for online sites that aim
to promote themselves as trustworthy entities to shop from: do they provide high
quality products that last a long time? Are they great at packing the items they ship?
Reputation mechanisms designed to judge the capability of trustees are based on
signaling [ 18 , 19 ]. In such cases, the more information there is about a specific item
of interest, the more informed the user will feel. This is the reason that most online