The Effect of Advice Message Location on User Performance - Professional Communications, IEEE Transactions on

In 1993, a study was begun with a large commercial oil and gas software interpretation system to determine the feasibility of a general taxonomy of on-line help content and a corresponding taxonomy of human interface access methods to this content. The preliminary findings from this work were encouraging and indicate that a taxonomic approach makes it easy both for help providers to understand what information they need to supply, and for help users to find the help they need quickly. Part of this taxonomy of help content includes inside application messaging. Existing studies of on-line help messaging indicate that both user-initiated or system-initiated advice messaging can improve user efficiency by prompting users with information about what something on the interface is, what it does, or what to do once it has been activated. This study examines the placement of on-line help messages in a multiwindow software application on user performance. Subjects were automatically timed as they performed two sets of tasks: one where help messages always appeared at the bottom of an application’s main window and one where help messages appeared at the bottom of the current window in focus (either the application’s main window or a secondary popup window). Half of the subject group were shown one message location first: the other half were shown the second message location first. The results demonstrated that, while the order in which the subjects were provided the two messaging schemes was not significant, users performed significantly better when the messages were placed in the current window in focus, and that this performance decreased as the size of a secondary window with current focus increased. Further, a majority of users showed a strong preference for help messages located in the current window in focus. AN the placement of on-line help improve user perforC mance? This was the question we asked ourselves in 1993, when studies to determine the taxonomy of on-line help content showed that a relatively high percentage of help information requested by users could be displayed inside the application. A review of on-line help literature reveals an expanding set of theoretical and practical answers to many of the design questions software developers are now asking about help. One striking thing about this research is how little of it deals first with what content end users of software applications actually require and, then and only then, with what access methods should be bound to what kinds of help [I]. Manuscript received May 1995; revised September 1995. S. G. McLellan and A. W. Roesler are with Schlumberger Engineering A. L. Elliot is at Georgia Institute of Technology, Atlanta GA 30332 USA. Publisher Item Identifier S 0361-1434(96)02285-0. Center, Austin, TX 78720-0015 USA. In 1993, a project was launched with a large commercial oil and gas software interpretation system to determine the feasibility of a general taxonomy of on-line help content and a corresponding taxonomy of human interface access methods to this content. The preliminary findings from this work were encouraging, and indicate that a taxonomic approach makes it easy both for help providers to understand what information they need to supply and for help users to find the help they need quickly [2]. Included in the taxonomy for on-line help were the following two categories: Information that is user-directed and answers the question “what is thislwhat does this do” about a component in the application’s human interface. User-directed help means that the user determines and initiates the kind of help desired-for example, by directing a cursor or keyboard focus on a particular interface element and activating help. Information that is application-directed and answers the question “what do I do next” about what operations are available once a particular interface element-for example, a button-has been activated. Application-directed means that the application determines and initiates the kind of help needed. The original study was divided into three separate experiments. Experiment 1 provided a set of questions asked by users needing help on a specific application. The experiment took the form of a recorded Wizard-of-Oz study, whereby a tutor (Wizard) received and answered all questions for help from subjects performing the same set of prescribed tasks on an application [3]. From the questions produced in Experiment 1, Experiment 2 showed that a substantial percentage of these recorded questions for help could be classified by help providers, using a proposed taxonomy, as either “what is” or “what’s next” questions (41%). This result was confirmed by Experiment 3, a repeat of Experiment 1 except for the substitution of an implemented help system for the Wizard. Here, too, 38% of the help required by subjects fell into these two categories of help. Both “what is” and “what’s next” categories of help had been implemented as messages displayed always at the bottom of the application’s main X window interface. Since the percentage of help information that was requested by subjects and that could be displayed inside the application itself was substantial, and since users needed help on components of secondary dialog boxes as well as the primary window, it made 0361-1434/96$05.00