The design of the user interface for an augmentative and alternative communication (AAC) or computer access system is a critical factor in determining a user's performance with a system. A comprehensive, quantitative, and accurate model for alternative access systems is needed to optimize both developers' design decisions and clinicians' system recommendations. This paper presents an application of one possible model, called the GOMS (Goals, Operators, Methods, Selection Rules) model (Card, Moran, & Newell, 1983). The model provides a comprehensive description of user behavior based on system-specific parameters as well as the cognitive, perceptual, and motor capabilities of the user. It can be used to predict both task execution and learning times, as well as points of excessive long or short term memory load. The GOMS model is applied here to three interfaces currently used in AAC and computer access systems in order describe and predict user performance, both qualitatively and quantitatively. The three interfaces are : (a) row-column letter scanning; (b) row-column letter scanning combined with word prediction after the first two letter selections only; and (c) row-column letter scanning combined with word prediction after each letter selection. Techniques for applying the GOMS model are discussed, as well as the results predicted by the model. Results for the three systems modeled here suggest the possibility that word prediction interfaces, developed as a faster alternative to row-column letter scanning, may actually be less efficient than letter scanning. KEY WORDS: alternative access, computer access, computer access for handicapped people, user performance modelingThe personal computer has tremendous potential for improving the functional abilities of individuals with physical and cognitive disabilities. Some of this potential has already been realized, and many new educational, vocational, and recreational opportunities have opened up for individuals with disabilities through the use of the computer.For a computer to be useful to individuals with disabilities, alternatives to the computer's hardware or software must often be developed. For example, a user who cannot physically use the standard keyboard must have an alternative means of accessing the computer, referred to as a computer access system. In addition, use of the computer as an augmentative and alternative communication aid for people who cannot speak requires a special user interface design, similar to that of a computer access system. Throughout this paper, both types of systems, that is, computer access systems with and without voice output, will be referred to as AAC (augmentative and alternative communication) systems.The user interface of an AAC system provides the user with a way to interact with the system, both in entering new information into the system (user input) and receiving feedback from the system (system output). The quality of this interaction is a determining factor in user performance, so the design of the user interface is...

No abstract

Unless these factors are correct, theoretic models will provide data which are neither relatively nor absolutely valid. If they are not soundly based on a series of very detailed studies of the behavior of actual users, GOMS models are unlikely to produce results that will be useful in any practical situation, particularly those involving disabled users.On the basis of our experience of disabled people using predictive systems, we have produced a simple description of one possible interaction technique, which shows an improvement in communication rate when prediction is used. We are well aware that we have left out many subtleties, and that there are other ways in which predictors may be used, but, unlike Horstmann and Levine (1990), the results of our analyses have been found to be similar to those obtained in trials with a number of users.We would not claim that prediction is a universal panacea, but we, and many therapists and teachers, have found that predictive techniques can offer substantial help to a wide range of client groups in terms of significant acceleration of text input and/or reduction in fatigue. They can also offer other advantages, such as reduction in cognitive load, motivation of users, and improvements in the quality of the written output (Newell, Arnott, et al., 1992; Newell, Booth, et al., 1992; Newell, Booth, & Beattie, 1991).We would not suggest that theoretic modeling is never appropriate, but we do believe that the lack of sophistication of the current GOMS modeling technique and our limited understanding of the detailed abilities of severely physically and cognitively disabled users means that it is not yet appropriate for detailed models of the GOMS type to be used as predictors of performance of users of AAC systems.It would be particularly unfortunate if those therapists with little practical experience of using predictive systems read Horstmann and Levine (1990) and, on the basis of the models and results therein, decided that there was little point in investigating whether the use of a predictive system could improve text entry for their clients.

No abstract

No abstract