Adaptive personalized interfaces—A question of viability

Abstract It is widely accepted that interfaces between computers and users should differ to accommodate individual, or group, needs. One method of 'personalizing' an interface is to have the system form a limited model of the user and employ it to fashion the dialogue to his needs. Unfortunately, little is known about the effect of adaptation on the man–machine interface. Although obvious advantages accrue from 'personalized' interfaces, there are also obvious disadvantages to presenting users with a changing, adapting and perhaps apparently inconsistent interface. The goal of this work is to determine the viability of an adaptive interface through a human–factor pilot study of a simple, specially designed, interactive computer system. The system uses menu–driven selection to retrieve entries from a large ordered telephone directory. This simple task has several advantages: it is a realistic application area for interactive computers; plausible adaptive modelling methods exist and have been studied theoretically; and previous work has determined the best way to display the menus to users. The results of this empirical study support the use of adaptive user modelling. In the (admittedly highly constrained) example system, a computer interface can indeed adapt successfully to every user. Although it does not necessarily generalize to other user interfaces, the result supplies evidence to refute published objections to adaptive user modelling in general.