choose to assess the credibility of the information herself in her evaluation of trust.
When trusting another for actions, she has to decide whether or not to trust, before
the action takes place. From a modeling perspective, this distinction can be easily
captured by a dependence relation. The trustor is dependent on one or more trustees
to varying degrees for her trust decision. One of the trustees can be herself to some
degree, based on her familiarity with the problem domain. Modeling the complex
dependence relations is a second important step in modeling trust. We will talk about
dependence in more detail in Sect. 5.2 .
We then outlined many different research threads that discuss how the various
trust constructs are evaluated. Instead of looking at trust as a single type of belief,
the trust context can outline the relevant beliefs that play a role in the evaluation
of complex goals. In particular, these beliefs may impact each other and evolve
differently. This is true even for computational agents that operate in complex
environments, such as the co-robotics example we mentioned earlier. We discuss
these issues in Sect. 5.3 .
Finally, there are many additional environmental factors that can impact trust.
We will review some of these in Sect. 5.4 . We propose to view trust as a continuous
process, in which the trustor's impression of the trustees evolves as new evidence
becomes available. This is true at different time scales, from cognitive processing of
signals of different complexity to changes in the underlying goals as a function of
newly discovered evidence. The modeling challenge in this case is identifying the
factors that can have a significant impact in the trust processing.
These four basic elements unwrap the notion of “trust context” across the large
number of domains that we have overviewed. We find that these elements have been
discussed in the literature to various degrees, but have not been brought together
under a single umbrella. As socio-technological networks drive the development
of applications with more complex trust contexts and new interdependencies, it is
increasingly important to study trust in a holistic manner. Without models of trust
that make explicit the assumptions underlying different trust computations drawing
from many different disciplines and contexts, it is hard to achieve this task.
The following sections outline the main modeling problems that must be
addressed for the study of trust in networks, discussed in decreasing order of
generality, from the most general network-level concepts to the details of trust
evaluation of individual subgoals. The different sections are part of a continuum.
There are many open research problems in each area and that provide many
interesting challenges to the study of trust in networks.
The Network View: Institutions Related
Throughout this brief, we discussed many institutions that take part in trust
decisions. When talking about Alice's trust for Bob, it is equally important to
consider Bob's trust for Alice. Alice and Bob come together in networks that provide