Cognitive task graphs and executable user models for better hypertext

Authoring hypertext systems is complex and difficult. To help designers build quality hypertext systems, designers should understand users’ behaviour and the common tasks they perform when navigating hypertext systems, and, further, designers should build in accurate representations of user models. This is uncontentious wisdom, but so far, not easy. This paper shows that common tasks can be broken down into simpler subtasks, taking into consideration users’ reasoning and learning. The results of this cognitive task-based approach can then be used to design the interface and functionality of a hypertext system. In particular we show that task trees can be automated, leading to ‘executable user models.’ Executable user models can be used efficiently as substitute users to support design and evaluation purposes. We discuss the task browsing as a concrete example.