Letizia is a user interface agent that assists in web browsing by tracking user behavior and anticipating items of interest through autonomous exploration of links. It uses heuristics to model user browsing behavior and provides recommendations for further action. The agent operates alongside a conventional web browser, such as Netscape or Mosaic, and can display recommendations for the user to follow or return to normal browsing. Letizia's design is based on the idea that information retrieval and filtering are intertwined, and it aims to recommend links that are most likely to satisfy the user's needs. The agent does not take control of the user interface but provides suggestions based on the user's browsing behavior. Letizia uses a strategy that is midway between conventional information retrieval and filtering. It leverages the user's browsing behavior to anticipate their interests and provides recommendations during idle time. The agent's recommendations are dynamically recomputed based on the current browsing state and user requests. Letizia's heuristics are based on user actions, such as saving a document, following links, or revisiting documents, which indicate interest. The agent also considers the context of the current document and the user's history to determine the relevance of links. Letizia's design is influenced by behavior-based interface agents, which rely on inferences from user actions rather than preprogrammed knowledge. The agent uses an extensible object-oriented architecture to incorporate new heuristics for determining interest in documents. Letizia's recommendations are based on a preference ordering of links rather than an abstract measure of interest. The agent aims to recommend a certain percentage of available links, set by the user. Letizia can also explain why it has chosen a particular document, based on factors such as keywords from previous explorations or comparisons to other documents. The agent's recommendations are valuable in uncovering serendipitous connections, which is a major goal of information browsing. Letizia's search strategy is a breadth-first search, which helps users explore links more efficiently by reminding them of neighboring links and avoiding dead-end links. Letizia's search is limited by resource constraints, such as a maximum number of accesses to non-local web nodes per minute. The agent does not initiate searches when it reaches a page with a search form, as there is no agreed-upon convention for time-bounding search efforts. However, it recommends that the user go to a page with a search form. Letizia's implementation is in Macintosh Common Lisp, using Netscape as the web browser and user interface. Communication between Lisp and Netscape is done via AppleEvents and AppleScript.Letizia is a user interface agent that assists in web browsing by tracking user behavior and anticipating items of interest through autonomous exploration of links. It uses heuristics to model user browsing behavior and provides recommendations for further action. The agent operates alongside a conventional web browser, such as Netscape or Mosaic, and can display recommendations for the user to follow or return to normal browsing. Letizia's design is based on the idea that information retrieval and filtering are intertwined, and it aims to recommend links that are most likely to satisfy the user's needs. The agent does not take control of the user interface but provides suggestions based on the user's browsing behavior. Letizia uses a strategy that is midway between conventional information retrieval and filtering. It leverages the user's browsing behavior to anticipate their interests and provides recommendations during idle time. The agent's recommendations are dynamically recomputed based on the current browsing state and user requests. Letizia's heuristics are based on user actions, such as saving a document, following links, or revisiting documents, which indicate interest. The agent also considers the context of the current document and the user's history to determine the relevance of links. Letizia's design is influenced by behavior-based interface agents, which rely on inferences from user actions rather than preprogrammed knowledge. The agent uses an extensible object-oriented architecture to incorporate new heuristics for determining interest in documents. Letizia's recommendations are based on a preference ordering of links rather than an abstract measure of interest. The agent aims to recommend a certain percentage of available links, set by the user. Letizia can also explain why it has chosen a particular document, based on factors such as keywords from previous explorations or comparisons to other documents. The agent's recommendations are valuable in uncovering serendipitous connections, which is a major goal of information browsing. Letizia's search strategy is a breadth-first search, which helps users explore links more efficiently by reminding them of neighboring links and avoiding dead-end links. Letizia's search is limited by resource constraints, such as a maximum number of accesses to non-local web nodes per minute. The agent does not initiate searches when it reaches a page with a search form, as there is no agreed-upon convention for time-bounding search efforts. However, it recommends that the user go to a page with a search form. Letizia's implementation is in Macintosh Common Lisp, using Netscape as the web browser and user interface. Communication between Lisp and Netscape is done via AppleEvents and AppleScript.