USER MODELING IN INTERACTIVE COMPUTER SYSTEMS

Abstract

This thesis discusses user modeling in adaptive interfaces in detail; presents a taxonomy of user modeling; identifies the dominant issues; describes an application amenable to adaptation; and investigates, through human factors experiments, the viability of adaptive systems. The term "user modeling" is examined first. A framework for characterizing adaptive interfaces is proposed, first through the presentation of two possible system architectures, and then through a taxonomy which classifies modeling into three types. The literature on fundamental issues in adaptive interfaces is then surveyed, but little is found in the way of empirical studies. In particular, the most fundamental issue - whether or not adaptive user modeling is a viable alternative to non-adaptive systems - remained unanswered. Repetitively accessed data bases, an application area amenable to personalization, are defined and examined through case studies. A personalized telephone directory system, built on an alphabetically-ordered menu interface, is constructed following guidelines derived from this work. The directory system is used as a test bed for examining the viability of adaptive interfaces. The results suggest that adaptive systems - at least in the personalized directory - can be far superior to non-adaptive interfaces, thus refuting many of the arguments found in the literature against personalization.