Reference: Walton, J. & Shortliffe, E. H. Automated Design of a User Interface. November, 1989.
Abstract: Making user interfaces effective for a variety of users is a recurrent problem in interface design. A custom-built interface may be effective, but editing it may require explicit program changes and it may be highly specific to the original user community. If a new group of users subsequently expresses interest in the system, the interface may not be appropriate to their environment or preferences, and it must be reprogrammed if the total system is to be successful in the new setting. The effort involved in reprogramming may be great enough to discourage dissemination of the system beyond its original environment. A possible solution to this problem is to provide an automated mechanism for adapting the user interface to fit the needs of the alternate users. A user-interface management system (UIMS) is one approach to automation that helps designers with the design, implementation, and maintenance of user interfaces. Our goal has been to create a prototype UIMS that produces the specification of a user interface for an application based on known design principles and on knowledge of the specific domain. Under this model, a dialogue with a potential user of the system acquires information about the user's preferences and environment. An inferencing system then combines this information with knowledge about user-interface design principles and builds the customized interface. This solution has been applied to the problem of accessing standardized patient medical records by different physicians in a variety of clinical environments. All physicians, regardless of their specialty, require access to the patient's full record. However, depending upon person preferences or the requirements of their specialty, they may want certain data to be displayed preferentially, emphasized over others, or put to different uses. To build a custom-tailored user interface for a prototype medical record, our system, called Tailor, interviews physicians about their use of patient records and dynamically builds personalized interfaces specific to those needs.
Notes: 42 pages.