Article Preview
TopIntroduction
Recently, a significant stream of information systems research has focused on developing and evaluating guidelines for conceptual modeling in information systems development. Much of that work has focused on justifying and evaluating prescriptions derived from ontology (e.g., Burton-Jones & Meso, 2006; Burton-Jones & Weber, 1999; Evermann & Wand, 2006; Gemino & Wand, 2005; Parsons & Wand, 2008). The primary research methodology used to evaluate ontologically derived modeling guidelines has been laboratory experimentation. Most prior work has evaluated the usefulness of modeling rules or guidelines by comparing the differences between following versus not following the rules, on measures of recall, comprehension, and problem-solving among readers of diagrams (Gemino & Wand, 2004). In this paper, we propose a complementary approach for assessing modeling rules. Specifically, we propose to examine the feasibility of automating the enforcement of modeling rules in software and to highlight design decisions that need to be made in the course of incorporating rules into a modeling tool. This kind of evaluation can also provide insights on the consistency and potential usefulness of such rules.
We evaluate a set of ontologically grounded modeling rules proposed by Evermann and Wand (2001a, 2001b) and Evermann (2003) to constrain the use of the Unified Modeling Language (UML) for conceptual modeling. Although the origins and initial focus of the UML lie in the area of object-oriented software design, it has evolved into a language for conceptual modeling – representing aspects of a real-world domain for which a system is required (Evermann, 2003; Burton-Jones & Meso, 2006). In this context, the UML is used as a language for communicating between clients and developers in understanding and eliciting requirements, and also for documenting the outcome of analysis (Dobing & Parsons, 2006, 2008).
The UML was developed as a language for software design (Booch et al., 2005). As a result, extending UML for use in conceptual modeling has raised questions about its appropriateness for that purpose. More generally, the applicability of object-oriented modeling in the early development phases is controversial (Kobryn, 2002). In addition, the UML in its entirety is very complex (Siau & Cao, 2001). In speaking of earlier versions of UML, Kobryn noted that “although UML 1.x has enjoyed widespread acceptance, its shortcomings include: excessive size and gratuitous complexity” (Kobryn, 2002). UML 2.0 and later versions have increased this level of complexity.
Against this backdrop, several researchers have proposed approaches to constraining the use of UML to conform to certain underlying principles. Such work is grounded in the view that, since conceptual modeling involves representing aspects of the real world, ontology, the branch of philosophy dealing with the nature and structure of the real world, is an appropriate foundation to guide UML-based conceptual modeling. Accordingly, researchers have proposed ontological rules that would constrain the construction of UML diagrams to ensure they properly reflect underlying ontological assumptions. If followed, these constraints effectively impose a method for using the UML for conceptual modeling that reduces the degree of freedom available to an analyst in creating a model. The outcome of following this method should be a model that is sound with respect to the ontological foundation on which the rules are based.