Modeling Web Navigation: Methods and Challenges

A computational cognitive model of Web navigation is a working computer system that simulates human users searching for items in a Web site. A fully working model must automate aspects of human perception, decision making and physical control. To successfully predict human behavior, these automated processes must be consistent with the cognitive and physical limitations of human users. Predicted behavior might include which links users select, when they select them and when they backtrack to previous pages. In this chapter, the necessary capabilities of a working model are described in detail. These include processes that simulate users scanning a page of links, assessing each link, selecting a link and deciding when to backtrack. Accurately modeling link assessment for a variety of users is critical for successful predictions and is perhaps the greatest challenge in creating a useful model. Several approaches to link assessment are presented. The implementation details of one model are described, which are then evaluated by correlating the model’s timing predictions to results from user studies.