Using menus to learn

Abstract

Transitional interfaces are based on the premise that if vocabulary and syntax are made accessible to new users they will be able to accomplish routine tasks as they learn about the system. Experiments conducted with UnixTutor (a transitional interface to the UNIX operating system) indicate that a training interface must do more than support vocabulary and syntax if it is actually to shield a new user from the vagaries of his environment. This is especially important because of the damage done by well-reasoned failures. In normal learning situations there is usually indication of approximations to success. This kind of feedback is rarely available from computers. Instead, the error messages and unmatched intentions continue until an exactly correct string of symbols is discovered. In the absence of an indication of approximate success this can result in a random walk among functional models because it is difficult to distinguish substantially correct models from substantially incorrect ones. It is concluded that neither the structured guidance of transitional interfaces nor the paternalistic blocking of training wheels offers a solution to this problem.<>