Many have noted the need for software to assist people in locating information
on the world wide web. This paper
presents the initial design and
implementation of an agent called WebWatcher that is intended to assist users
both by interactively advising them as they traverse web links in search of
information, and by searching autonomously on their behalf. In interactive
mode, WebWatcher acts as a learning apprentice [Mitchell et al., 1985,Mitchell et. al., 1994],
providing interactive advice to the Mosaic user regarding which hyperlinks to
follow next, then learning by observing the user's reaction to this advice as
well as the eventual success or failure of the user's actions. The initial
implementation of WebWatcher provides only this interactive mode, and it does
not yet possess sufficient knowledge to give widely useful search advice. In
this paper we present WebWatcher as a case study in the design of web-based
learning agents for information retrieval. We focus in particular on the
interface that enables WebWatcher to observe and advise any consenting user
browsing any location on the web, and on results of initial experiments with
its learning methods.