Letizia is a user interface agent designed to assist users in browsing the World Wide Web. It operates in tandem with a conventional web browser like Mosaic or Netscape, tracking the user's browsing behavior and making recommendations to enhance the browsing experience. The agent uses a set of heuristics to model the user's interests based on their browsing actions, such as following links, initiating searches, and requesting help. Upon request, Letizia can display a page with its current recommendations, which the user can choose to follow or ignore. The core of Letizia's design is its control structure, allowing users to manually browse documents and conduct searches without interruption. Letizia's role is to observe and infer the user's interests from their actions, making suggestions when the user is unsure of what to do next. It employs a best-first search strategy, augmented by heuristics, to explore links and predict the user's future needs. This approach interleaves both information retrieval and filtering, adapting to the user's browsing behavior and preferences. Letizia's effectiveness lies in its ability to leverage the user's "idle time" while reading documents, making recommendations without interrupting the browsing process. It can also remember and maintain the user's interests over time, capturing persistent interests and serendipitous connections. The agent uses a breadth-first search to compensate for the depth-first orientation of many web browsers, ensuring that users don't get lost in hyperspace. The implementation of Letizia is done in Macintosh Common Lisp, using Netscape as the web browser and user interface. Communication between Lisp and Netscape is facilitated through AppleEvents and AppleScript interprocess communication. The project has been supported by various research grants from institutions like Alenia, ARPA/JNIDS, Apple Computer, and the National Science Foundation.Letizia is a user interface agent designed to assist users in browsing the World Wide Web. It operates in tandem with a conventional web browser like Mosaic or Netscape, tracking the user's browsing behavior and making recommendations to enhance the browsing experience. The agent uses a set of heuristics to model the user's interests based on their browsing actions, such as following links, initiating searches, and requesting help. Upon request, Letizia can display a page with its current recommendations, which the user can choose to follow or ignore. The core of Letizia's design is its control structure, allowing users to manually browse documents and conduct searches without interruption. Letizia's role is to observe and infer the user's interests from their actions, making suggestions when the user is unsure of what to do next. It employs a best-first search strategy, augmented by heuristics, to explore links and predict the user's future needs. This approach interleaves both information retrieval and filtering, adapting to the user's browsing behavior and preferences. Letizia's effectiveness lies in its ability to leverage the user's "idle time" while reading documents, making recommendations without interrupting the browsing process. It can also remember and maintain the user's interests over time, capturing persistent interests and serendipitous connections. The agent uses a breadth-first search to compensate for the depth-first orientation of many web browsers, ensuring that users don't get lost in hyperspace. The implementation of Letizia is done in Macintosh Common Lisp, using Netscape as the web browser and user interface. Communication between Lisp and Netscape is facilitated through AppleEvents and AppleScript interprocess communication. The project has been supported by various research grants from institutions like Alenia, ARPA/JNIDS, Apple Computer, and the National Science Foundation.