Consistency is one of the important measures to evaluate the usability of user interfaces. TAG(Task Action Grammar) model tries to represent the degree of consistency of user's task knowledge based on rule schemas. It is often not proper, however, to access consistency solely by considering syntactic structure of active operations ignoring the responses by the system.
System responses may be crucial in user interfaces since users will not always rely on the memorized complete action sequence for a task but may utilize system responses to determine what to do next. This suggests that model must be extended to include the system response that indicates the system states.
When the user's task is to change some attributes of work-objects in a system, the system response should be perceived and interpreted in relation with the particular work-object. Therefore, from the user's point of view, the system state is most eminently characterized by the state of the work-objects. The system response, together with the user actions for a task, must be taken as the most important element of a model to describe the user's task knowledge.
This study conducted experiments to uncover how the consistency of system responses affects user performance. Two systems are used in the experiment. The two have the same degree of consistency of action sequences from TAG's point of view, but differ in consistency in terms of the system responses. Experimental results show that the rule schemas that consider the system responses can more accurately represent consistency of the systems. This suggest that the model of TAG may be extended to take the system responses into account to provide better indication of interface consistency and the resulting user performance.
