Hence, even for the same problem, the dependence relation is not a static
distribution of weights. The familiarity (or expertise) of the trustor in a specific
domain is often a factor that has an impact on the weights. If the trustor has
knowledge in a domain, then she can judge the credibility of information herself by
relying on herself. When she is not capable of judging information herself, she may
then rely more heavily on the reliability or expertise of the information sources for
her trust decision. The corroboration of the information by many sources may play
an important factor, especially to reduce the dependence on information sources that
are not trusted. Hence, the trustor may seek out more information in domains where
she is not an expert and does not have access to known and trusted sources. Also, the
credibility of the information can be a significant factor in judging information trust.
The credibility of information can be also based incorrectly on familiarity, instead
of expertise. Information that is easy to understand and recall may be considered
more credible. For example, if Alice is seeking an answer to a puzzle question, she
may find an incorrect answer more credible than the correct answer if the former
sounds more familiar than the latter.
These weights are also affected by the cognitive or other computational resources
needed to compute the trust decision. Even though the trustor is capable of reviewing
information, she may decide to rely on the source's trustworthiness in many cases,
simply because it is easier to do so. This is especially true if the source is highly
trusted. In that case, relying on the source does not carry any risk and hence there is
little reason to judge the credibility of the information. Note that it is possible to find
algorithmic equivalents of this type of approach that rely on specific trusted nodes
for a computation, instead of aggregating over a network.
Hence, the second type of design decision in modeling trust involves determining
the important subgoals for a trust decision and how they are combined. The trust
context also defines how the weights of different subgoals change what conditions
they depend on. Often these conditions correspond to how much information the
trustor has for a specific subgoal. As a result, this is an appropriate place to model
the uncertainty inherent in trust decisions as a function of the available information
for evaluating trust.
Trust Constructs and Beliefs
Trust constructs associated with each individual subgoal build on the dependency
relations. In the previous section, we talked about different subgoals and how the
dependence on each subgoal may change contextually. This section concentrates on
a single subgoal and how it is evaluated by further refining the above model.
When evaluating trust for another entity, we outlined two main dimensions that
are common to many different threads of research: trustworthiness and compe-
tence. The trustworthiness construct describes the integrity, reliability, and positive
intentions of the trustee. The competence construct describes the ability of the