Reference: Sherman, E. H. A User-Adaptable Interface to Predict Users' Needs. Elsevier Science Publishers B.V.,, 1990.
Abstract: User Modeling has become an important and common approach for designing user interfaces (UIs). Many different types of user models, both quantitative and qualitative, have been proposed. Some of these models have been integrated into guidelines, and some have become the basis of standard techniques for UI design. But using these techniques can be both time consuming and costly, particularly if an application is expected to be used by multiple user communities, with different patterns of use. A UI designer must choose between spending a great deal of time and money designing an interface for each distinct user community, and designing a less ideal, single, generalized UI for all user communities. User-interface management systems (UIMSs), a set of tools to support UI design and implementation, have been used to speed and improve the process of UI design. They were originally built to assist the traditional UI designers, computer scientist. They also have been used to make it easier to adopt user models in the design of UIs and to automate the integration of user models into UIs. UIMSs have allowed experts in human-factors design, without experience in computer programming, to create UIs. More recently, UIMSs have been created to incorporate users knowledge into the design of a UI automatically. We have created a UIMS called PODIUM (personalized designer: an intelligent user-interface manager). The purpose of PODIUM is to show that a UIMS that is supplied with a UI automatically designed for a large and diverse user community can tailor that UI automatically for each of many subgroups of a large user community. The only knowledge that the system uses to complete this task is the user characteristics that divide the community into subgroups and its experience with users who have tailored UI's previously. PODIUM was built for the domain of the physician's clinic chart and was tested on physicians from two specialties. The physicians were given the opportunity to edit their UI interactively, and these experiences were used by PODIUM to guide the building of UI's for other physicians of the same specialty. Two different UIs resulted from this procedure, one for each specialty. The physician users evaluated the feasibility and usefulness of the process, revealing that UIs built by PODIUM when it used more experience were indeed preferred to these PODIUM built using less experience.
Notes: Updated November 1993.