Page 78 - Trust in ICT 2017
P. 78
1 Trust in ICT
Figure 22 – A Trust Data Usage Ontology
8.3 Development of a static policy/rule-based trust-level decision making mechanism
Trust models with decision making mechanism are trust models that provide both (i) rules, formulas and
algorithms describing how to compute trust, and also (ii) hints on how to use that information in the decision
making processes. The evaluation protocol and the used metrics differ, depending on what the decision
making mechanism does.
While the trust evaluation phase has been extensively studied, approaches for decision making mechanism
often employ very simple models. Often, the agent who is ‘most trusted’ is automatically selected for
delegation, without considering any other factors. Risks, rewards, and the potential for trustees to make
deliberate choices, are often not considered.
8.3.1 Specify policy/rule for deciding trust levels
Once trust evaluations have been produced for a given set of individuals, the decision to trust must be made.
This problem has been approached in different ways by some existing trust models, and neglected entirely
by others.
The trust policy is used by the trustor as well as trust platform to define the diversity of personal preferences
that they wish to impose on their perspectives of trust. There are many possible policies depending on the
context, trust model and infrastructures.
Here are some trust policy and rules perspective depending on the trust model for decision-making
mechanisms:
• Cognitive View: This cognitive approach explicitly considers the inseparable nature of trust, risk and
context.
While trust in another individual may be higher than for any other, the trustor may stand to lose too
much to make delegation preferable. On the other hand, the trustor may have so little to lose and
so much to gain, that he is willing to consider even those partners who are not especially
trustworthy.
The cognitive approach argues the need to keep separate the process by which an agent forms trust
beliefs, and the process by which an agent decides to act on trust, by delegating. While the cognitive
view is abstract and far richer than any existing computational model, the authors show that the
different trust beliefs can be reduced to a single degree of trust suitable for use within a decision-
theoretic framework.
• Exploration and Thresholds: The trustors who possess utility functions for each attribute of a
service, and these are used when evaluating services after an interaction. Agents can, therefore,
define a threshold of utility, here co-operation may be considered. As this is not a probabilistic
model, this utility cannot be considered ‘expected’ in the decision-theoretic sense. In their
evaluation, consumer agents are initially randomly distributed in the environment and have a
preference for interacting with agents who are ‘nearby’ in the environment.
70