encapsulated by a utility function, the perceived desirability of a situation according
to the trustor. A positive consequence is considered to have positive utility. Similarly,
negative consequences are determined by their negative utility. The desirability of
different situations is subjective and may change from trustor to trustor. Various
methods to formalize these preconditions of trust have been introduced. We
summarize these conditions following conventions chosen by Wagner et al. [ 19 ]:
1. The trustor has more than one option to choose from. The trust decision is made
before the trustee acts.
2. The outcome in the trusting option
depends on the trustee
. In the non-trusting
, the outcome does not depend on the trustee.
3. The trustor is better off if she chooses
and the trustee ends up
being trustworthy, given by a utility function
4. The trustor is worse off if she chooses
and the trustee ends up
being not trustworthy.
The first condition corresponds to our definition of options. The second condition
describes the notion of dependence. The final two conditions describe the notion of
In our example, if Chip fails in his duty, Alan will fall and badly injure himself.
This situation clearly has low or negative utility. If Chip does as he is expected,
Alan will finish his job successfully, a desirable condition with positive utility. If
Alan does not trust Chip, other options will not produce a nice coat of paint and
lead to lower utility than trusting Chip.
We sum up the above options by the equation of expected utility
^ x D
^ x D
It has long been argued whether the above elements can describe what trust is.
The above are conditions that exist in all situations requiring trust: the presence
of choices, the dependence on the trustee in the trusting choice, the uncertainty
and the vulnerability of the trustor at the decision time. However, these conditions
are not necessarily complete since they abstract out elements like the common un-
derstanding of expectations between the trustor and the trustee. Many sophisticated
computational trust models have been developed with the aim of understanding how
people trust others or for building agents that can trust another entity.
Trust is affected by the trustor's knowledge of her environment, as perceived by
her. The environment provides her with input relevant to trust where the relevance is
determined by her goals. Exactly how the environment and the trustor's knowledge
impacts trust is typically part of the trust model implicitly. However, we will make
this aspect of trust models explicit by our emphasis on the trust context in a wide
range of applications. We will not concentrate on a single formalism for modeling
trust, but instead we will describe how contextual elements change the trust models.
Next, we describe the crucial elements of trust context.