Context-centric proactive information delivery

Abstract

Delivering only those information relevant to the needs of teammates is critical for avoiding information overload. Need-driven proactive information delivery has thus been widely investigated in diverse areas spanning from psychology, AI, to Web-based services. CAST, a team-oriented agent architecture, realized proactive information delivery as a built-in agent behavior, which is derived by the capability of anticipating others' information-needs based on a computational shared mental model. However, the first implementation of CAST is limited in managing the dynamic activation and deactivation of teammates' information-needs. This work addresses this issue by proposing a mechanism for organizing and managing the "context" of information-needs, which facilitates information-needs to be activated and deactivated dynamically. We evaluate the proposed approach by comparing its performance with agent teams using only reactive communications (e.g., ask/reply) and with the original CAST agent teams.