A knowledge-based approach to design intelligent interfaces
Hikmet Senay · 1987
Modern computer systems continue to provide sophisticated software tools which perform more and more complex tasks that are known little to most users. To bridge the gap and facilitate the use of such sophisticated software by a growing user community, computer systems also provide on-line help facilities. Even though the important role of help systems in user-computer interaction is well-recognized, on-line help still remains as one of the most neglected issues in many interface designs. Based on field studies of actual help system use, this dissertation identifies several difficulties encountered in interface to help systems. It then describes a knowledge-based interface model which provides system- and user-initiated, context-dependent, user-specific, and adaptive help. In essence, the help system is an integral part of user interface to all other software tools. Within this framework, command language is viewed as a fuzzy set whose boundaries change in time and are different for each user. In accordance with this view, a fuzzy parsing algorithm is developed to map user inputs into the command language. An essential component of the interface is a user modeling subsystem which uses a stochastic learning automaton that acquires knowledge about user experience with domain concepts. This knowledge is utilized during parsing as well as help presentation. The interface is implemented in Prolog to take advantage of its nondeterministic behaviour. The prototype implemented is evaluated in a controlled experiment which shows that the interface model reduces the effect of errors significantly and has potential to be more usable and user satisfying.