The Web is a vast, dynamic source of information and resources. Because of its size and diversity, it is increasingly likely that if the information one seeks is not already there, it will be soon. Unfortunately, finding the right places to look, and persistently revisiting those places until the information is available is an onerous task. In this paper, we describe Do-I-Care (DICA), an agent that uses both technical and social mechanisms to ease the burden of locating "interesting" new information and resources on the Web.

DICA monitors Web pages previously found by the agent's user to be relevant for any changes. It then compares these changes against a user model, classifies them as potentially interesting or not, and reports the interesting changes to the user. The user model is derived by accepting relevance feedback on changes previously found. Because the agent focuses on changes to known pages rather than discovering new pages, we increase the likelihood that the information found will be interesting.

DICA combines an effortless collaboration mechanism with the natural incentives for individual users to maintain and train their own agents. Simply by pointing DICA agents atother agents, changes and opinions canbepropagated from agent to agent automatically. Thus, individuals train and use DICA for themselves, but by using a simple technical mechanism, other users can use those results without the additional effort that often accompanies collaboration.