information source. Hence, trust is not a fixed value between two entities. It is a
function of the trustor, the trustee and the context. What is in this context?
To understand the elements of context, we need to examine what factors impact
the trust relationship between two entities. Going back to the example of whether
Alice trusts Siri to find a doctor, Alice's goal is to find a good doctor. The criterion
considered in determining trust could be the expertise of Siri on the given topic, i.e.,
health. The reason that Alice does not trust Siri for choosing a doctor could be that
she does not yet think of Siri as an expert in the topic of health. However, Alice's
needs in this topic may go beyond expertise. Does Siri know enough about what
Alice looks for when searching for a good doctor? Maybe Alice cares greatly about
the doctor's bedside manner. Can Siri account for this appropriately?
An important factor in deciding to choose Siri for such a recommendation could
be the underlying intent behind Siri's operation. Is Siri built to help people or to
promote certain information sources based on monetary considerations? In other
words, does Alice believe that Siri has good intentions towards its user? Alice may
not have firsthand information to arrive at a decision on this topic and may instead
consider it as an Apple product. She might love Apple and trust that Apple products
are built with users' needs in mind. As a result, she might believe in Siri's good
intentions as an Apple product.
Alice's goal may be more complex than just getting a doctor recommendation.
She might not have a problem with getting doctor recommendations from Siri as
a way of generating an initial list and then sifting through them herself. She also
thinks that she can send some names to her friends and get their opinion. The
contextual component of a decision involving trust incorporates both the goals
and the dependencies inherent in these goals: Alice's dependence on the intelligent
agent, on the entity/company that produced it, on herself and on her friends.
Most trust modeling work does not explicitly mention the context of a trust
evaluation because it is generally implied by the specific domain the work is
originating from. While some complex models implicitly incorporate some elements
of context, there is no explicit effort in this area. As we will see, different theories
may apply when considering dependence on different entities and different goals.
An Example of Different Trust Contexts
The issue of context is quickly becoming very important due to increasingly
sophisticated networks that combine social systems with computational systems. As
a running example, we consider the newly emerging field of co-robotics (Fig. 2.2 ),
which is dedicated to the development and use of robots that work beside, or
cooperatively with, people. This field provides us with many interesting examples
of trust from very different contexts. All of these different types of trust must be
addresed in the design of the new generation of robots.
Consider a robot that is designed to help people construct buildings by providing
physical support and help. How can the person trust the robot?