Résumé en anglais

The growing number of services available within social applications (viz. Social networks) raises a new and challenging search issue: selecting desired services from social networks. Traditional discovery and selection approaches, which are registry-based (e.g., UDDI, ebXML), have manifested their limitations as they often fall behind users' expectations. This is because registries fail to (i) take into consideration non functional properties such as QoS and trust and (ii) capitalize on the information resulting from the previous experiences between agents. To address these shortcomings, we use software agents as they support interactions and offer well-developed capabilities to formally express and interpret semantic information useful to evaluate trust. Trust in a service is a multi-aspect concept that includes a social-based aspect such as judging whether the provider is worthwhile pursuing before using his services (viz. Trust in sociability), expert-based aspect such as estimating whether the service behaves well and as expected (viz. Trust in expertise) and, recommender-based aspect such as assessing whether an agent is reliable and we can rely on its recommendations (viz. Trust in recommendation).