Human-computer interaction research benefits from the scientific method as much as any other field and it should be theory-driven to the greatest degree possible in order to make significant advancements. The leading researchers and their corresponding important contributions have all followed this path, and the quality of their work is a testimony to the success of this approach (e.g., Card, Moran & Newell 1983; Guiard 1987; Shneiderman 1998; Tullis 1997). As Landauer ( 1997) pointed out, there are at least 3 goals for research in the field of HCI. First, research is required in order to understand what system to build and what features and functions to design for which users and tasks. Secondly, we need to perform research to develop better scientific theories, models and principles about user behavior during computing tasks. Finally, the goal of obtaining high quality guidelines or standards for user interface design that can be handed off to designers and developers requires formal research. But how do we go about performing these badly needed areas of study? According to Landauer, research in HCI differs to varying degrees from the scientific processes and methods used in such areas as psychology. Specifically with regard to formal lab studies, significance testing and control, the ideal cannot always be attained in HCI research. In addition, the design process does not always lend itself to systematically testing one variable at a time. This perhaps should not be so surprising, since the discipline of HCI uses mostly an engineering process model, and pulls from such varied areas of expertise as Computer Science, AI, Linguistics, and Psychology, among others ( Preece 1994). In what follows I will briefly attempt to formalize a scientific approach to research in HCI that can be practically used in the design of user interfaces now and in the future.
It is extremely important that we follow a formal scientific process in HCI in order to move our domain expertise forward and communicate broadly amongst