The authors wish to thank Patricia Everaert (Ghent University) and two anonymous reviewers of the 2 nd Int’l Workshop on Conceptual Modeling Quality (IWCMQ’03) for their useful comments and suggestions on previous versions of this paper. D/2004/7012/05

Interest in evaluating conceptual modelling techniques has recently experienced a revival, in part due to widespread adoption of the Unified Modelling Language (UML). In addition, the use of ontology as a framework for evaluating conceptual modelling techniques has gained acceptance. In this paper, we consider implications of applying one aspect of the ontology of Mario Bunge to conceptual modelling. Specifically, conceptual modelling has traditionally failed to provide mechanisms to indicate that some properties of types or classes may be considered dependent on others. This paper presents a theoretical rationale, using Bunges ontological notion of precedence, for explicitly modelling such dependence in conceptual schema diagrams. We present the design of an experimental framework to test the impact of explicitly representing precedence on the ease with which a diagram can convey domain semantics. In addition, we consider how the issue of common sense semantics can interfere with experimental procedures to evaluate the semantics conveyed in a diagrams structure. We offer early experimental results indicating: 1) the explicit modelling of precedence improves the ability of experimental participants to verify the existence of dependence among properties (but has no effect on the ability to verify the semantics conveyed by association cardinalities); and 2) the potential for background knowledge to interfere with the semantics conveyed by diagram structure. We conclude by discussing the need for further research on both these issues. Keywords: property precedence, ontology, experimental evaluation of conceptual modelling techniques. Copyright 2004, Australian Computer Society, Inc. This paper appeared at the First Asia-Pacific Conf...

Abstract. This paper presents a laboratory experiment that evaluates the REA approach for modelling enterprise-wide accounting information systems. REA is a pattern-driven conceptual modelling approach that is based on the Resource-Event-Agent semantic model of a company’s accountability infrastructure. Using a between-subjects experiment with business students we investigated whether Entity-Relationship (ER) diagrams that show a REA pattern occurrence are better understood than informationally equivalent ER diagrams that do not show a REA pattern occurrence. The results of our experiment indicate that students develop a more accurate understanding of the business processes and policies modelled when they recognize the REA structure of accounting information in ER diagrams. Students also perceive such diagrams to be easier to use when performing model comprehension and validation tasks. The experiment further showed that the number of information systems courses taken by students has a significant effect on task performance. The number of accounting courses completed before the experiment did not affect the understanding of business process models. The paper concludes by discussing some implications for Accounting Information Systems