TOOL USE, REUSE, AND ORGANIZATION IN COMMAND-DRIVEN INTERFACES

Abstract

To handle the diverse implements they wield in their workshops, people group tools into functional and task-oriented collections, and set recently-used ones aside for re-use. Surprisingly, these strategies have not been transferred effectively to interactive computer interfaces. The chief problem in designing a command interpreter that lets people reuse and organize their on-line activities is the dearth of knowledge of how users behave when issuing commands to general-purpose computer systems. Consequently, existing user support facilities are \fIad hoc\fR designs that do not really support natural work habits. Recent studies which examined people's behavior in interactive interfaces paid undue attention to command choice and not enough to complete command lines. By examining both aspects, this thesis abstracts general principles governing how often people repeat their activities from usage data gleaned from different classes of user over several months. These provide design guidelines for "history" mechanisms that make old submissions available for re-use. The problem is to identify likely candidates, and several ways of conditioning the distribution to enhance predictive power are evaluated. A case study of actual usage of a widely-available history system is included. Users also organize their activities by task and by function. This can be supported by an on-line "workspace" that allows people to group tools for related activities. A system loosely based on the metaphor of a handyman's workbench is described and used to illustrate the problems that are encountered when facilities to expedite interaction are bolted on to existing computer systems.