This paper describes how to generate compositions of semantic Web services using social trust information from user ratings of the services. We present a taxonomy of features, such as interoperability, availability, privacy, security, and others. We describe a way to compute social trust in OWL-S style semantic Web services. Our formalism exploits the users’ ratings of the services and execution characteristics of those services. We describe our service-composition algorithm, called Trusty, that is based on this formalism. We discuss the formal properties of Trusty and our implementation of the algorithm. We present our experiments in which we compared Trusty with SHOP2, a well-known AI planning algorithm that has been successfully used for OWL-S style service composition. Our results demonstrate that Trusty generates more trustworthy compositions than SHOP2. 1 Motivations Web services are finding life in new forms, which is forecasting the future of what the web service environmen...