Information provided by a source should be assessed by an intelligent agent on the basis of several criteria: most notably, its content and the trust one has in its source. In turn, the observed quality of information should feed back on the assessment of its source, and such feedback should intelligently distribute among different features of the source-e.g., competence and sincerity. We propose a formal framework in which trust is not treated as a monolithic and static concept. We regard trust as a multi-dimensional concept relativized to the sincerity of the source and its competence with respect to specific domains: both these aspects influence the assessment of the information, and also determine a feedback on the trustworthiness degree of its source. We provide a framework to describe the combined effects of competence and sincerity on the perceived quality of information. We focus on the feedback dynamics from information quality to source evaluation, highlighting the role that uncertainty reduction, and social comparison play in determining the amount and the distribution of feedback.