Human-computer interaction : from classifying users to classifying users' misunderstandings

The overall objective of the research has been to address the question of how best to understand user behaviour at the interface. The use of cognitive grammars to analyse tasks and predict behaviour was rejected for seven theoretical and practical reasons. Following this, cognitive style measures were rejected as a result the first study, where the visualizer-verbalizer and conceptual tempoc ognitive style measures were not found to be accurate predictors of behaviour at a task. The results of this experiment indicated that interaction between a system and its user has certain dynamic qualities that make prediction of a fixed set of activities in a set order difficult. Furthermore, it seemed likely that behaviour is determined by a potentially complex interaction of variables rather than any single over-riding factor, such as a user's cognitive style. Consequently, attention was-focused upon the errors that occur during humancomputerinteraction. An approach where errors are classified was -adopted, and a classification scheme was developed (ECM: an Evaluative Classification of Mismatch)as a vehicle for further research. An initial pilot study showed that user-system errors could be classified using the scheme. This suggested that the concepts it employed did have some validity in'both cognitive and computing domains. The second study of ECM involved a design team at Hewlett Packard's Office Products division in Wokingham. This study demonstrated that the classification scheme was - usable by a design and development team that consisted of software engineers, human factors engineers, and technical authors. The third and final study of ECM demonstrated that it could be, used to improve a design. A system, that had been changed using ECM, was shown to be significantly better, in terms of time, errors and user attitude ratings, than either its original or an iteration where ECM had not been employed. This research has provided strong indications that evaluative classifications can be of use within the design and development process. Furthermore, this work emphasizes the importance of providing structures for thinking about the user's problems that are divorced from the structure and terminology of design.