A test-bed for user interface designs

Most presently available interactive computer interfaces treat their users in an unfriendly, uncooperative, and inflexible way, resulting in feelings of frustration and a conseqeuent loss of productivity for the users. These problems have led to attempts (e.g. 6, 8, 12, 13) to make interfaces appear more friendly and cooperative through the addition of advanced interface features such as spelling correction, on-line help, personalized defaults, etc.. While common-sense suggests such features may be helpful, there is little hard evidence about how helpful they are or whether they are worth the overheads they entail. A primary reason for this lack of information is the practical difficulty of experimentation. Many of these features are time-consuming to implement, are usually implemented without adequate instrumentation, and are implemented in different and difficult to compare ways from system to system (see 10, for example). These problems in evaluation suggest the need for a test-bed interface in which various advanced features could be tried out in a consistent and adequately instrumented way with a variety of application systems. In this paper, we present a detailed rationale and a partially implemented design for a test-bed of this kind.