The Impact of Experience on Individual Performance and Workload Differences Using Object-Oriented and Process-Oriented Systems Analysis Techniques

Interest in object-otiented analysis (OOA) and object-oriented design (OOD) has increased over the last several years. Proponents of OOA and OOD have called the shift to these methods “revolutionary” and have cited a number of impressive claims with respect to their use. However, empirical research investigating these claims remains in its infancy. Moreover, an examination of the literature reveals inconclusive and often conflicting results for OOA and OOD methods relative to traditional data anoYorprocedura1 techniques. This research study was conducted to test theoretical hypotheses designed to better understand the findings of previous research. i%is paper presents the results of an empin’cal study which examined both experienced and novice systems analysts using both procedural and object-oriented techniques. The study used a 2X2X2 repeated measures partial factorial design with expen’ence and method order as between subject variables and analysis method as the within subjects variable. This research examined whether experience in using procedural methodr helped or hindered pelf0 rmance and OOA methods on the subjective mental workload induced. using OOA and compared procedural A cognitive processing model adapted from DeSanctis (1984) and Huchins, Hohan, and Norman (1985) provided the theoretical framework for this study. attitudinal measures. Dependent variables include subjective mental workload, time to peflorm task, and While some of these dependent variables have been used in past research, we believe this is the first t ime that subjective mental workload (&WV) (common in the human factors literature) has been examined in an IS context. Our results indicate that both novice and experienced subjects demonstrate higher SMW when using OOA, however, as a group, novices prefer object-oriented techniques and find these techniques easier as compared to experienced subjects. Introduction between user and analyst arefundamentally dtjj%rent with OOA than with structured analysis.” Fichman and Kemerer (1992) conclude that the shift to object-oriented approaches represent a radical change from previous approaches to software development. Therefore, it seems logical to expect that the impact of this “radical” paradigm shift would have important ramifications for systems analysts. Unfortunately, there is a surprising lack of empirical research investigating many of the claims outlined above. Thus, it remains unclear whether the use of object-oriented approaches have many of the desirable impacts suggested by proponents of OOA. Some researchers have shown that object-oriented methodologies improve the completeness of solutions while reducing complexity associated with using the technique (Boehm-Davis and Ross, 1992). However, others have found object-oriented methodologies more difficult to apply and harder to learn than process-oriented approaches (Vessey and Conger, 1994) or data-oriented methods (Bock and Ryan, 1993). Thus, the limited empirical research conducted to date has yielded equivocal results. In addition, given practical and methodological constraints, most of the research to date Interest in object-oriented analysis (OOA) and objectoriented design (OOD) has snowballed over the past few years. Proponents of OOA and OOD have proclaimed a number of benefits for use of these techniques. For example, some researchers have suggested that objectoriented approaches might assist developers by easing the process of design and improving the usability of the resulting system (Rosson and Alpert, 1990). Others suggest the advantages of OOA and OOD are due to their ability to model complex relationships (Coad and Yourdon, 1991). In general, many researchers consider the primary advantages of OOA and OOD to be in their ability to reduce complexity. Given these benefits, some have called object-oriented approaches a new “paradigm” for software development (Lee and Pennington, 1994). Many have gone so far as to call this “paradigm shift” revolutionary (e.g.Booch, 1989). Yet another group of leading OOA and OOD proponents, Coad and Yourdon (VW), claim (emphasis added) “the discovery process, and the communication 1060-3425196 $5.0001996IEEE 232 Proceedings of the 1996 Hawaii International Conference on System Sciences (HICSS-29) 1060-3425/96 $10.00 © 1996 IEEE Proceedings of the 29th Annual Hawaii International Conference on System Sciences 1996 has used extremely limited sample sizes (e.g., 6 in Vessey and Conger, 18 in Boehm-Davis and Ross). As a result, it appears that additional theory-based, empirical examinations of different systems analysis approaches are appropriate. Given the claims about the “radical” nature of the shift to object-oriented methodologies, we wanted to examine whether analyst experience plays a role in making this shift. On one hand, one could argue that experience in performing systems analysis tasks would “carry over” from process methods to object methods allowing experienced analysts to outperform novices. On the other hand, if the paradigm shift is as radical as has been suggested, one could suggest that systems analysts experienced in using conventional methods (e.g., data flow diagrams) would find it difficult to transition to a revolutionary way of analyzing problems. In either case, these effects would seem to have important ramifications for the personnel assignments, training, and hiring of systems analysts. In order to examine the influence of systems analysis experience on the use of OOA and process-oriented techniques, a laboratory experiment was conducted. Subjects were drawn from two different populations: one inexperienced in systems analysis and design and the other experienced in systems analysis and design using a process-oriented method (data flow diagrams). A repeated measures design was employed to compare a processoriented method (data flow diagrams) with an objectoriented approach (a variation of Coad and Yourdon notation) for each group. Dependent measures included subjective mental workload, time to complete the solution, perceived ease of method application, and method preference.