A Fuzzy Implementation for the Socio-Cognitive Approach to Trust - Trust Theory: A Socio-Cognitive and Computational Model - page 321

1) . At the beginning of the experiment these values are randomly assigned to each agent on
each possible task.
The Willingness represents how much the agent will be involved in performing tasks (e.g. how
many resources, will or amount of time it will use); this modulates the global performance
of the agent in the sense that even a very skilled agent can fail if it does not use enough
resources. Each agent has a single willingness value that is the same for all the tasks it tries
to perform; it is a real number that ranges in (0, 1) .
The Delegation strategy is the rule an agent uses for choosing which agent to delegate
the task to (e.g. random, cognitive, statistical). It is the variable we want to control in the
experiments for evaluating which trustor performs better.

Agents reside in an environment that changes and makes the tasks harder or simpler to
perform. Changes are specific for each agent and for each task: in a given moment, some
agents can be in a favorable environment for a given task, some others in an unfavorable one.
For example, two different agents, performing the same task, could be differently influenced by
the same environment; or, the same agent performing different tasks in the same environment
could be differently influenced by it in performing the different tasks. Influences range in ( 1,

1) for each agent for each task; they are fixed at random for each simulation. The environment
changes randomly during the simulations: this simulates the fact that agents can move and the
environment can change. However, for all experiments, if a task is delegated in an environment,
it will be performed in the same one.

−

11.14.2 Delegation Strategies
In the contract net, on the basis of the offers of the other agents, each agent decides to whom
to delegate (Castelfranchi and Falcone, 1998) depending on their delegation strategy. We have
implemented a number of different agents, having different delegation strategies :

a random trustor : who randomly chooses the trustee to whom to delegate the task. This
kind of trustor has no a priori knowledge about: the other agents, the environment in which
they operate, their previous performances. There is no learning. This is used as a base line.
a statistical trustor : inspired by a number of works, including (Jonker and Treur, 1999),
assigns a major role to learning from direct interaction. They build the trustworthiness of
other agents only on the basis of their previous performances, without considering specific
features of these agents and without considering the environment in which they performed.
It is one of the most important cases of trust attribution; it uses the previous experience
of each agent with the different trustees (failures and successes) by attributing to them
a degree of trustworthiness that will be used to select the trustee in a future interaction.
There is a training phase during which this kind of trustor learns the trustworthiness of each
agent through a mean value of their performances (number of failures and successes) on the
different tasks in the different environments; during the experimental phase the statistical
trustor delegates the most trustful agent (and continues learning, too). There is no trustor's
ability to distinguish how the properties of the trustee or the environment may influence the
final performance.