Search in Referral Networks

Referral systems have been proposed to assist people in finding potential experts in person-to-person social networks, in which each user is assigned a software agent and software agents help automate the process through a series of referrals. However, most of the existing referral systems have focused on the referral information generation, and simply consider referrals through path search in a static graph. An important question is how to efficiently search the social networks with the help of software agents, while agents only rely on their local knowledge. In this paper referral networks are proposed to model the social structure emerging among software agents. We study the referral networks empirically for the community of AI scientists (as defined in bibliographic data), and show how to control the searching process by adaptively choosing the referrals.