Discussion, Conclusion, and Future Research

The results presented here focus on two main dimensions: performance, as measured by online time spent on the task, and satisfaction, as measured by a system satisfaction scale.

To start, performance was generally better overall for experienced users (103.12), as opposed to novices (141.70), and the difference was significant. However there was a large and considerable difference between users of the command language (177.55) versus the wizard (67.27). Also of interest is the fact that the wizard tended to equalize performance across skill levels; in other words, both novice and experienced users spent roughly the same amount of time to complete the task. It is of course expected that experienced users would complete the task in less time than novices, and the wizard would bring about faster performance. However, it would have seemed logical that experienced users would have still complete the task faster than novices using the wizard, but that did not turn out to be true.

From the perspective of system satisfaction, there was no significant difference between novice and experienced users (F = .497, p = .481). However, there was a difference when it came to system satisfaction as it related to interaction mode (F = 29.01, p = .000). In general, wizard users were far more satisfied overall than command language users (17.21 versus 21.55).

In addition, it appears there is a greater difference in satisfaction between interaction modes for novices than for experienced users. Clearly, users with experience in programming and markups would find the interaction mode to be less critical in terms of their satisfaction with the system - simply different ways of doing the same thing. However, novices may find the wizard approach to be user-friendly and enjoyable, while the experience of using the command language and editor may prove to be rather trying and difficult.

In general, it could be concluded that if the goal is to complete a task quickly, the wizard is definitely the mode to use. Both experienced users and novices were able to complete the task significantly faster than using the command mode. However, if command language coding is used, experienced users were able to complete it in twice the time needed, compared with the wizard. In the case of novices, it took approximately three times as long.

Experienced users were roughly equally satisfied regardless of the interaction mode used, a reflection of their previous experience in programming and markup languages. Novices, on the other hand, were quite affected by interaction mode, expressing great satisfaction with the wizard and far lower levels of satisfaction with the command language mode.

System designers who are looking to design programming or Web development systems should apply these findings to systems they are looking to design or enhance. For tasks where prompt execution and fast completion is desired, the best choice would be to offer a wizard feature so that users of any skill/experience level could use it easily. The addition of the wizard feature could help to enhance acceptance of the system and encourage users to work with it frequently. On the other hand, it appears that while novices greatly appreciate wizard-based systems, experienced users do not show the tendency to the same degree, and it is possible that some experienced users would actually prefer using the command language interaction. Further research which examines the relationships between these interaction modes and learning, as well as the effects upon other measures of performance, could yield more insights into the impacts of interaction mode, on using and learning markup languages.

There are a number of research questions and issues, which are deserving of further research in command languages. These include the following areas:

  • Combined menu/command language approaches. The combination of menu techniques together with traditional command language approaches is worthy of further study, especially in the case of learning command languages. How can command languages be presented to help someone learn the language more quickly and effectively?

  • Formatting of commands to conform to certain metaphors. When teaching a command language, what kinds of formats and organization structure can be emphasized to enable someone to more easily learn the language? Can they be designed to fit specific metaphors or structures?

  • New methods of organizing command languages to make command languages easier.

  • There are a number of areas relating to menus/form fill-in where further research and study would be both interesting and fruitful. These are in the areas of:

    • Nonlinear menus. These are menus which are not laid out in a traditional linear organization. Instead, menus can be patterned using a more meaningful ordering of items. For instance, a menu can be arranged in a tabular format, as a circular form as on a clock, or in a pie, geographical map, or color wheel formats. Do these alternate formats improve performance, satisfaction, and usability?

    • Interdependent menus. Menus can be designed so there are relationships and dependencies between items. Instead of selecting "a, b, c, d," there can be menus that follow an organizational chart, process/function, or "tree" structure. The structure of the interdependent menu can be defined in a graphical display instead of a linear list.

    • Analog menus. These menus, using pointing devices, allow for the entry of continuous values.

    • Improving the performance on "deep menus." Most research points to broad and shallow menus are more efficient than deep ones. Are there methods of improving performance on narrow, deep menu structures?

These are just some of the issues and questions, which can be explored in further research of menus.

This study produced a number of useful and meaningful results in better understanding human-computer interaction as it relates to markup languages. There are a number of interesting and meaningful research directions, which can be explored with this study, as a starting point. They include logical extensions to the research conducted here, including the use of different kinds of subjects, or perhaps, doing this kind of study on different kinds on new languages which are currently being developed and used, especially for the Internet and Web development.

Infrequent/Casual Users and Markup Languages. There is a body of literature (Martin & Fuerst, 1987; Trumbly, 1988), which examines the behavior and performance of intermittent users, such as doctors, lawyers, and other kinds of professionals. They use computers on an irregular, intermittent basis, and are neither true novices nor true experienced, using computing systems. While they likely would not perform well on an involved programming application (like C or C++), it would be useful to study their performance, learning, and satisfaction on a task such as using markup languages. In fact, many professionals probably would want to understand how to build Web and Websites for their businesses and practices. In a future study, where subjects of this type are available, this would be an interesting investigation to pursue.

Adaptive Interfaces for Novices and Experienced. Another extension of this research, which would require greater resources for advanced programming and coding of the experimental software, would be to create an innovative kind of "adaptive" interface which would be able to adjust to the level of the subject in providing the proper interface and help systems for the user.

Dynamic/Adaptive Menus are menus which "adapt to the current user by compensating for weaknesses, providing help appropriate to the context, and decreasing the mental and physical workload of the particular user" or more simply put, should conform to the idea that "the interface should adapt to the user rather than the user adapt to the system." The system should be able to provide appropriate help when necessary, compensate for user weaknesses, as well as decrease the mental load placed onto the user (Norcio & Stanley, 1989).

There are two main types of adaptive interfaces. This includes both dynamic adaptation by the system, as well as interfaces which can be modified by the user. The first type is what is generally referred to as an adaptive interface. Some of the options available for adaptive interfaces include user selectable and definable options/choices, dynamic menus, which adapt to the user automatically, or moving the cursor to the next needed item rather than at the beginning of the menu. These options can be run at the user's desire and discretion, or made continuously running. The important issues involved in adaptive menus, or adaptive interfaces in general, include the user models, which support these kinds of systems, as well as the dialogue between the system and its users (Norcio & Stanley, 1989).

There is a debate concerning whether these kinds of adaptive menu interfaces are desirable. Those who support it cite the user's knowledge of the task, the need for different interfaces for different people, reduction in learning time, user satisfaction, and reducing mental workload. Opponents claim static interfaces are better, because the user prefers to learn and use one interaction style, makes skills portable across systems, reduces learning time due to a standardized arrangement, reduces cost of implementation, creates the feeling of a "loss of control," as well as results in the inability for a user to develop a model of the system due to constant changes (Norcio & Stanley, 1989; Mitchell & Shneiderman, 1989).

Adaptive interfaces, while simple to define, are not simple to build, as it requires a knowledge base, which consists of four types of domain knowledge. These follow those classifications suggested by Rissland (1984) and Croft (1984): knowledge of the user, knowledge of the interaction, knowledge of the task/domain, and knowledge of the system.

For instance, knowledge of the user is important for adaptive interfaces, especially in the creation of a "user model." A user model is "the description and knowledge of the user maintained by the system." The user model in an adaptive interface varies according to user, and is modified by the system if the user changes. What factors are important when designing a user model for an adaptive system? Potosnak (1984) suggests that they might be computer experience, computer knowledge, and program-specific knowledge. The differentiation between "novice" and "expert," based on levels of experience, is commonly used. Of course, even classifying users, using this system is difficult. How can novices and experts be differentiated? This might be the types of commands used, number of times on the system, times requesting help, or something else. A problem with the "novice-expert" set of user models is that when a user reaches the point where the system is directed to upgrade him or her to "expert" status, then the changes made can be disturbing to the user. Errors, declines in performance, and excessive help requests can all come about, when the system is changed to "expert mode." Instead, the process should be a continuous and gradual one (Norcio & Stanley, 1989; Maskery, 1984).

Aside from this, there are various other variables that can be considered. Some of these include differences in information processing skills, cognitive styles, interaction preferences and styles, as well as spatial and verbal abilities. Yallow (1980) did some research in this area and found the amount of retention of material was based upon the individuals' low/high spatial ability, and the format of the information provided (verbal vs. graphic). If the abilities of users are known, then it is easier to create specific modules, either to remedy or to maximize the individual abilities involved. Another example concerns the use of windowing in an adaptive system. Since people differ in their attention capabilities, and windowing allows for multiple windows of information, tasks, and programs at the same time, it seems appropriate that different windowing layouts would be required to assure that all information displayed is received (Robertson, 1985). The user's mental model of the system is also important (Norcio & Stanley, 1989).

Knowledge of the interaction and dialogue is also important, especially if natural language is used. A knowledge of the domain and task is also important, and the creation of models for the task, goals, and plans (Norcio & Stanley, 1989).

An experiment with a dynamic menu that changed its structure to place the most frequently used commands always at the top was tested against a static menu system. The results showed that users tended to learn the structure of a menu system rather quickly, and that the changes in order were disturbing. The dynamic menus slowed down novice users and, even after practice, dynamic menus did not bring about improved performance over the static arrangement.

The overall reaction to the dynamic menu was negative; however, giving users more control (change order, not change order, amount of order changed) over the changes might have improved satisfaction. In general, the results showed dynamic menus, although a good idea, may not be as effective as originally thought, although no definite conclusions can yet be drawn at this point (Mitchell & Shneiderman, 1989).

Despite the negative findings presented in the previous research, Norman and Chin (1989) call for further work into the creation of self-adapting menus, which create the optimal arrangement of the menu, based on what the user has selected before frequently, described by Berke and Vidal (1987), as the Most Frequent Use (MFU) metric. While this claims to close the gap between the designer's and the user's conceptions of the system, poorly designed adaptive menus can cause confusion, due to the changes in the arrangements of the menu items. User-adaptable menus (Chin, 1988) allow the use to change and arrange the existing menu structure (or names of the options), to make it more personalized and fit his or her interaction style more effectively. Hypercard pulldown tearoff menus are examples of this (Norman & Chin, 1989).

Possible functionality of the experimental software might be a kind of "pre-use questionnaire" which queries the user on his/her background and preferences, and then presents (and sets up) a suggested "customized" interface and help system for that user. Users could be tested in various ways: being forced to use this interface, and/or have the choice to choose the type of interface and help system he/she wants to use by making choices from the check boxes/radio buttons on the screen. The results obtained from the use of this "adaptive" form of screen would be interesting, and would extend the work which has been previous done in terms of adaptive interfaces.

Another option might be to have a longer experiment (perhaps a kind of repeated measures design) which asks novice users to start with a novice interface and help system, and then have him/her repeat the task again using an experienced interface, with the expectation that the second time around he/she would be more familiar with the system and task.

Survey Experts and Markup Languages for Surveys. One area, which can be examined in more detail, given a subject pool of marketing research professionals, survey designers, and other related experienced staff, would be to participate in an experiment such as this, to see if there are any significant impacts or effects for survey experts.

Experimental Testing of Other, New Internet/Web Languages. There are many other kinds of new languages, which are being introduced, especially for the Internet and Web development. In particular, while the results of this study can be generalized to all forms of markup languages, it would be interesting and instructive to see what results would come about in studying languages for Internet scripting, such as JavaScript, VBscript, or Perl. It would be interesting to examine user interaction in terms of learning, satisfaction, and performance, using languages such as these. The continually advancing developments in the area of markup languages, including DHTML, VRML, HTML 4, and XML ensures there will be many opportunities for further research in the area of markup command languages.



Computing Information Technology. The Human Side
Computing Information Technology: The Human Side
ISBN: 1931777527
EAN: 2147483647
Year: 2003
Pages: 186

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net