"Conventional techniques for finding information on the web are time-consuming and often yield unreliable results. This has led to renewed interest in referral systems, which support the interactions of people with others by directing users toward others who might have the answers they seek. We consider a social network of agents who assist each other in helping their users find information. Unlike in most previous approaches, our architecture is fully distributed and includes agents who preserve the privacy and autonomy of their users. These agents learn models of each other in terms of expertise (ability to produce correct domain answers), and sociability [...]"