| Sumario: | The area of trust and reputation mechanisms for virtual societies is a recent discipline oriented to increase the reliability and performance of electronic communities by introducing in such communities these well known human social control mechanisms. This thesis has two different parts. In the first part we present ReGreT, a sophisticated trust and reputation system oriented to complex societies where the social component of the agents behaviour has a special relevance. The system uses the knowledge about the social structure of the society as a method to overcome the lack of direct experiences and to evaluate the credibility of witnesses. By combining direct experiences, third party information and social knowledge, the system can improve the computation of trust and reputation values. It also provides a degree of reliability for these values and can adapt to situations of partial information improving gradually its accuracy when new information becomes available. A trust and reputation system like ReGreT only has sense if it is used by a deliberative agent. The second part of the thesis is devoted to the use of multi-context systems as a means of specifying and implementing this kind of agents. We propose several extensions to the multi-context approach that facilitate the specification and implementation of complex deliberative agents. As the main example of using multi-context systems to specify deliberative agents we present the specification of an agent's component implementing the ReGreT system.
|
|---|
