A Fuzzy Implementation for the Socio-Cognitive Approach to Trust - Trust Theory: A Socio-Cognitive and Computational Model - page 331

600 600 600 600

373

100

Figure 11.15 Experiment 7: real time, introducing the first trustworthy strategy (measuring time spent).
(Reproduced with kind permission of Springer Science+Business Media C 2005)

of the task; he can decide for example to stop the delegation and to change delegee. In this
case, the agent has a new attribute: controllability . A more complex task resolution scheme
and cost model (e.g. the costs of stopping or changing the delegation) is needed, too.
In order to exploit the controllability, we changed the task resolution scheme. In order to
be successful, an agent has, as usual, to score a certain number of hits (e.g. three). In order to
score a hit, instead of performing a number of attempts equal to their willingness, the agent
has exactly 10 attempts : for each attempt, he first tests his willingness (i.e. if it actually tries to
perform it); if this test is successful, it tests his ability (i.e. if the delegee is able to perform the
task). Each test is simply 'rolling a random real number in (0,1): if the roll is less than the tested
attribute (Willingness or Ability), the test is successful. The environment can interfere with
the agent's activity setting the Difficulty of each Ability test. At the end of the ten attempts, if
the number of scored hits (e.g. three) is sufficient, the task is performed; otherwise it is not.
In addition to the usual tests of willingness and ability, for each one of the ten attempts each
Agent checks if the delegee sends a report (a controllability check); the report contains the
number of the current attempt, information about the activity in this attempt (tried/not tried;
performed/not performed) as well as the number of successes achieved. At the end of the ten
attempts, the agent sends a final message that says whether the task was successful or not.
Monitoring consists of receiving and analyzing the reports. Basing on this information, an
agent can decide either to confirm or to retire the delegation: this can only be done before
the final message arrives. If they retire the delegation, they can repost it (only one more
time). Obviously, agents having a higher controllability send more reports and give more
opportunities to be controlled and stopped (and in this case the tasks can be reposted).