Characters on the Web


Mike Perkowitz
Dept. of Computer Science, University of Washington
with Gene Ball, Microsoft Research

The Web is a big place. Even an individual’s personal history and bookmarks quickly become unmanageably large. We take a social rather than entirely technological approach to this problem in the design of Argent, our browsing assistant. Argent is based on the idea of sharing Web references with friends and coworkers - that is, with others who have distinct and familiar personalities. Argent presents an interface to a number of characters, each with its own personality - various interests and tastes in Web pages. Like people, Argent’s characters maintain their own home pages, complete with self-portraits, self-descriptions, and links to favorite Web pages. Characters can suggest interesting pages to visit and links to follow, and a user can recommend a pages to characters.

A group of people - perhaps a workgroup or a circle of friends - can share a set of characters. Any member of the group may interact with these characters and make recommendations to them. Over time, the characters will come to represent a collective memory of the group’s adventures on the Web, organized and mediated by their personalities. These personalities develop in response to user recommendations and may shift significantly over time. The Argent model provides a user means of organizing her Web experience, of sharing references with others, and of discovering new pages. As a research tool, Argent also provides a testbed for exploring personality in interfaces, collaborative filtering, and the relationship between an individual, her peers, and the digital world.