Theory and Research Approach

Research and practice reports on UML are diverse and scattered (Cho & Kim, 2001, 2002;

Sim & Wright, 2001, 2002; Johnson, 2002). We find documentation of negative as well

as positive effects of UML deployment. Also, UML — while studied and described in

many computer science papers — has generally not been studied in the evaluative sense:

what is its economic contribution to a firm? How does the precision of description it

provides trade off against the time it takes to develop and maintain? What is its cost

through the lifetime of an application and for use throughout an organizational department?

Most profoundly, in our view the established literature typically addresses only

a small subset of UML-related issues within a very narrow area of the project effort.

Reflecting upon how UML is used in the organizational context, it quickly became

apparent that decisions and implementation at multiple levels must be involved. Clearly

the organization must permit, if not encourage, its use. Particular projects may subscribe

or not subscribe to the UML approach — perhaps for engineering reasons such as a

project whose mission is not deemed consistent with this approach or for human resource

reasons such as it would take too long to acquire UML development skills to allow the

project to be completed on time. We came firmly to believe that use of UML requires the

resolution of organizational, project, group, and individual issues. With the exception

of textbooks, contributions addressing UML in this larger context could not be found.

The challenge of conducting research on UML in an organizational context is further

discussed in the subsections of search for a theory, theory-driven versus data-driven

research, and use of causal mapping.

Search for a Theory

The lack of an integrative perspective in UML research may not be typical in research

addressing similar research objectives. A review of published research within the topic

area of management information systems identified diffusion theory as a likely candidate

(Fichman, 2000). If we consider UML as a “technology” in the broad sense of a technique

and approach to achieving a set of results, the way that it is diffused and adopted among

organizations, projects, groups, and individuals would perhaps be a reasonable framework

for attempting to address UML usage and the impact it has on project outcomes.

Rogers’ (1995) formulation, sometimes called Rogers’ theory, has been fairly thoroughly

examined using highly precise and rigorous measures — generally gathered through

widely distributed survey questionnaires (Brancheau & Wetherbe, 1990; Moore &

Benbasat, 1991; Prescott & Conger, 1995; Fichman & Kemerer, 1999). Rogers’ theory has

also been extensively criticized (Larsen, 2001; Lyytinen & Damsgaard, 2001). For

example, organizational IS/IT innovations are more complex than Rogers’ theory specifies.

The IS/IT innovation processes unfolding in the “adopting” organization are richer

and more diverse than sigmoidal. The division of users into the categories of early

adopters, early majority, late majority, and laggards is at best unproven but more likely

an introduction of social complexity, yet simplicity, that implies semantic meaning

contrary to innovation process realities (Van de Ven, Polley, Garud & Venkataraman,

1999). Decisions are not purely individual but distributed among individual actors,

groups, project managers, and organizational units. We concluded that using diffusion

theory as the platform for studying UML in an organizational context would be inappropriate.

Our concern parallels the debate that has appeared in venues ranging from top journals

to informal discussion groups about the “rigor” versus “relevance” and “theory” versus

“description” characteristics of research studies (for example, Robey & Markus, 1998;

Benbasat & Zmud, 1999; Nurminen & Eriksson, 1999; Goles & Hirscheim, 2001; Weber,

2003). Given how little knowledge links UML to organizational outcomes, the study of

UML would suggest “relevance” and “description.” At least among the organizations

we ended up visiting, the use of UML remains either a new technology that could

potentially be adopted or one that is partially in use with the jury still out regarding its

precise cost-benefit results. Therefore, adding to the knowledge of UML in the organization

would potentially be of benefit to these practitioners. In order to integrate our

observations regarding the research characteristics of UML with theories of diffusion,

we decided to take a theory building approach.

Theory-Driven vs. Data-Driven

Research

We found ourselves in a predicament — staying theoretically pure or reporting on the

use of object-oriented analysis and design as it unfolds in organizations today. We came

to believe that there is a dialectic struggle between theory and practice (Van de Ven &

Poole, 1995). The employment of theory requires a precise definition of constructs,

variables, relationships, and units of analysis, or has these elements as the planned

outcome of the research effort. Conversely, increasing our understanding of a complex

phenomenon, such as the use of object-oriented design and analysis, requires a

multilevel research design where the outcome cannot be predicted based on available

research reports (Klein & Kozlowski, 2000).

Our research focus on increasing our understanding of UML in its broad organizational

setting suggests that many loose ends and disconnected bridges would emerge. The

indication is that chaos rules. Although this might be expected, we also believe that some

degree of order will exist. In organizational settings, actors allocate time and resources

to issues and phenomena that they believe have importance. A high demand for

resources in defining and executing action related to the issues and phenomena

addressed, as is the case with UML, will increase the likelihood that actors take it

seriously. It is reasonable to suggest that actors will concentrate on particular issues and

phenomena that they deem as being important. Directing attention to issues and

phenomena is a necessary but not sufficient prerequisite for understanding why actors

would pay attention. We also believe that actors are concerned about outcome. That is,

when time is spent on defining issues and phenomena, attention would also be given to

the impact that one of these may have on other issues and phenomena.