her goal. The trust evaluation may also take into account the network context. For
example, the social network can be trusted to punish bad behavior and make it more
likely that Bob will perform a specific action. In this example, both Bob and the
network are trusted to some degree as part of the overall goal of the trustor. In short,
trust evaluations involve complex goals with possibly multiple trustees.
Different trust goals lead to different concerns and processing. While many
factors can be considered in evaluating trust goals, we will concentrate on two main
classes: the competence and the trustworthiness of the trustee. Most trust goals
can be mapped to either one or both of these classes of goals, which we will refer to
as trust constructs . When relying on Bob to hold a secret, his trustworthiness is the
most important concern. When relying on Bob to hold a heavy object, his physical
capability is more relevant. Informativeness of signals varies depending on which
trust construct is being considered. For example, trustworthiness is impacted by first
impressions, especially by those obtained from facial attributes. For competence,
first impressions may come from information related to the education and the
training of an individual. Information obtained from the network and the trustor's
firsthand experience with the trustee play a role for both constructs, but the trustor
may pay more attention to positive evidence for competence and negative evidence
for trustworthiness. Furthermore, one expects that competence is specific to a topic,
while trustworthiness is a more general belief that applies to a large number of goals.
In short, the goals and the underlying trust constructs determine which signals are
important for trust evaluation and how evidence is evaluated.
The trust evaluation differs for different types of trustees in various ways; judging
the trustworthiness of a person versus the trustworthiness of a robot may be quite
different in terms of human cognition. As a result, the trustee may have an impact
on how trust is evaluated.
Finally, external conditions such as the ordering of events and availability of
different signals may also change the result of trust evaluation for the same trustee(s)
and the same goal(s). This is especially true for human trustors. These dynamic
considerations are part of the trust evaluation environment , which describes the
conditions that impact the trust evaluation beyond the trustor, the trustee and the
trustor's goals. They can be considered as a part of the dynamic properties of trust
context. Note that these elements of context are not orthogonal categories; they are
inter-related in many ways.
In summary, the trust context is a system level description of trust evaluation
that takes as input the following variables/tunable parameters: (a) the trustor and
the network she is operating in, (b) her goal(s) in making a trust evaluation
and the underlying trust constructs, (c) the trustee(s) that she depends on, and
(d) the environmental conditions that impact the trust evaluation. Next, we inves-
tigate the impact of each variable in trust evaluation in detail. However, we first
describe the preconditions of trust that are common to all trust contexts.