Exploration and experienced performance with display-based systems

The research reported in this dissertation investigates how interface-literate users explore and use new display-based computer applications. It aims at identifying characteristics of present interface designs that help or hamper these activities. Task-oriented exploration of a new application is conceptualized as search through a problem space of legal interactions. Three interface characteristics were derived on the background of this conceptualization, and their combined effects on task performance were investigated. In the experiment, display-based interaction of experienced users was investigated for the case of graphing tasks with commercially available computer applications. Performance during task-oriented exploration and two further trials at two different delays was analyzed globally and on a very low level of detail. The overall results show that subjects were able to complete the exploration phase when help was provided. Knowledge acquired in the exploration phase could be used successfully later on. Specifically, this knowledge helped to prevent search behavior in a second trial. Forgetting effects were relatively small even after one week. Mechanisms for learning and possible causes for the rare occurrence of forgetting effects are discussed. The detailed analyses show evidence for the use of a search heuristic, called label-following (Engelbeck, 1986; Polson, 1988). Search through a problem space of legal interactions can be constrained by semantic matches between interface labels and the users' goal representation. Two design characteristics, the use of unlabeled direct-manipulation interactions, and the use of 'poor' labels, are shown to impair the use of the label-following heuristic, and therefore hinder discovery of the correct actions during exploration. A third design characteristic, the provision of large sets of possible operators, is shown to complicate search in these situations. These characteristics are shown to impair performance during exploration as well as after long delays. Possibilities for application of the results in guidelines for evaluation of display-based interfaces are discussed. Finally, the chosen empirical methodology is evaluated on the dimensions of generalizability, applicability, and usability.