تاثیر مدل های فیش آی آگاه از متن در درک فرآیندهای کسب و کار : مطالعه تجربی از نمودارهای جریان داده

We investigated whether a “context-aware” fisheye view can more successfully communicate the information contained in a set of process models (data flow diagrams) than a traditional “context-free” presentation. We conducted two controlled experiments: the first included a simple set of DFDs and tasks that required a basic understanding of the system, while the second involved more detailed views of the same processes, and also a more complex task. Subjects who used the fisheye process models outperformed those using the traditional presentations. This difference was reflected in task performance for all subjects, and in task completion time for inexperienced subjects.

Business processes were defined by Davenport and Short [8] as “a set of logically related tasks performed to achieve a desired business outcome;” they are the building blocks of a business system. Once those processes are understood, they can be improved or radically redesigned in order to improve an organization's productivity [18]. Thus representing processes by means of easily understandable models is essential. The process models to represent complex systems can also become complex. In such cases, multiple interrelated diagrams are often used [22]. Human capacity for simultaneously processing multiple pieces of information is limited [28] and [29]. Therefore, an effort to simultaneously explore all relevant business processes in detail would create a quantity of information that cannot be processed easily: an information overload. To avoid this, process models are simplified into levels, depending on the degree of detail. For models representing nontrivial systems, there will be several levels, and the viewers of these diagrams must integrate context and details while switching back and forth between levels. Such integration is not easy and may cause disorientation with viewers feeling lost. System analysts are therefore faced with a dilemma: one option is the creation of a single comprehensive model, while the other is the creation of separate models representing different parts of the overall system, possibly causing confusion when switching between diagrams at different levels of detail. According to Kim et al., this effect complicates the cognitive processes in integrating these diagrams to understand the system as a whole. This dilemma is not unique to process models. For example, while browsing hypertext, the user can become confused about the link structure between nodes and how one leads to another. This phenomenon has been called “lost in hyperspace” [6] and [31] and presentations of web sites that provide better integration of context and details have been shown to be more successful [5]. Similarly, Schafer et al. [33] found that context clues improved viewers’ performance when navigating and rerouting the links of a simulated telephone network. We suggest that process models would be more useful if they had “context-aware” features so that the diagrams were organized to integrate context and details smoothly. The fisheye view concept introduced by Furnas [10] seemed particularly promising. It has been applied to graphical presentations of hierarchies for groupware [16], hypertext [9], and search on the World Wide Web [41]. Turetken et al. [40] and Turetken and Schuff [39] proposed the use of fisheye views with systems analysis and design diagrams. Specifically, the data flow diagram (DFD) is a good process model for such a visualization technique because of its strong hierarchical structure and due to the fact that it is a classic tool for process analysis. Accordingly, our research question is: Can the presentation of Data Flow Diagrams (DFDs) through fisheye views convey the business processes that they represent more successfully than the standard presentation of such diagrams?

6.1. Overview Our study introduced fisheye DFDs as context-aware representations of an organization’s business processes. We created and tested two sets of DFDs, and found promise in their use in understanding business processes. In our first study, we found support for our hypothesis that the use of the fisheye views enabled subjects to understand business processes more completely. Our second study provided additional insights into fisheye process models and their effect on task score and completion time. First, fisheye models appear to be most effective at improving understanding when subjects are field dependent. Second, the fisheye models appear to be most effective at improving task completion time when subjects are inexperienced. This would imply that the fisheye models may be most effective with novice users. Both of our experiments were conducted with samples of moderate sizes. The fact that we found some highly significant results suggests that the effect size (i.e., the contribution of fisheye views to under- standing of the modeled system) is very large. The findings also suggest that it is possible to gain simultaneous performance and speed improvements; therefore it is reasonable to try to achieve both effects in visual interface development and testing. Qualitative evidence that was collected appeared to support these findings. Some subjects who were given the fisheye version responded with comments thatspecifically addressed the integrated nature of the diagrams. Some were: ‘‘It was helpful when the DFDs were exploded, to see how they stayed integrated with the rest of the system.’’ ‘‘The [ease] of being able to zoom in on areas gave a good overall picture of the whole system.’’ ‘‘It helped because it was easier to look ... at each process [in depth].’’ 6.2. Implications of the study 6.2.1. Implications for researchers This study contributed to our knowledge of systems analysis and visualization. One way to apply a gestalt approach in the study of systems is during its graphical modeling. We used context-aware visualization tech- niques to demonstrate a novel approach in the presentation of business processes by integrating process details and context within the same diagram using a ‘‘fisheye’’ view. Although the focus of this study was understanding and improvement of business processes, there are aspects of systems that can be modeled through fisheye views within a system design and development domain. For example, UML, with its ability to represent activity flows in an object-oriented application development environment, could be a viable application domain. The concepts of aggregation and inheritance have hierar- chies that may lend themselves to fisheye views. Also, our findings have implications for under- standing how individual differences affect the success- ful use of visualization techniques. In our second study, strong interaction effects between cognitive style and level of experience and presentation method highlight the need for usability studies to be attentive to the moderating effects of individual level differences on task success. This finding is particularly interesting in that it contradicts previous work that suggests that cognitive style is not a determinant of task-based performance [36] . Although we were not specifically interested in the main effects of these individual variables, their inclusion in our models revealed interaction effects that provided greater insight into the specific circumstances under which a certain visual presentation is more successful. These results provide further insight into other research [27] that advocated the use of different information views during the systems development process as a way of addressing users’ cognitive limitations. Our findings specifically addressed one such limitation caused by users’ inherentcharacteristics or internal factors as suggested by Barkin and Dickson. 6.2.2. Implications for managers and developers Managers of organizations that use graphical models that have a hierarchical nature should consider our findings. As the integration of disparate IS becomes prevalent, it is important to understand not only the system itself, but also its context. Hasselbring [20] , for example, observed that linking suppliers with custo- mers has extended the boundaries of their IS to include each other’s systems. Our findings also offer guidance for CASE tool developers. We used two variations on our user interface in this study. The first consisted of static web pages, and the second enhanced those images using animated transitions and color. Such visualizations could be integrated into CASE tools through an application modulethatdynamicallyaltersmulti-levelmodelswithin computer-based systems analysis and design tools. 6.2.3. Implications for educators The use of this visualization technique has sig- nificant implications for teaching systems analysis. The benefits from the fisheye view suggested that inexper- ienced people realized significant speed improvements when using our techniques. Therefore, the opportunity to present relationships between subprocesses, their parent processes, and the other processes in a system more effectively should be helpful in addressing the challenge of introducing these concepts to students.