Construction and Pre-Test of a Semantic Expressiveness Measure for Conceptual Models

One quality attributed to McCarthy’s Resources-Events-Agents (REA) accounting model is semantic expressiveness. Compared to accounting models without a semantic orientation like the Debit-Credit-Account (DCA) model, the REA model is claimed to better represent the economic phenomena underlying an accounting system. The alleged benefits of this increased semantic expressiveness include easier integration with representations of non-accounting information and better user understanding of accounting systems. Dunn and Grabski (2000) have empirically compared the semantic expressiveness of accounting systems based on the REA and DCA models. They showed that accounting students perceived the REA system as more semantically expressive. Furthermore, they demonstrated that higher perceived semantic expressiveness (PSE) leads to higher accuracy in accounting information retrieval tasks. To measure the PSE of a system they used a single 7-point Likert scale assertion, which was applied after performing an information retrieval task. As ideas for future research they suggested the development of better PSE measures, as well as investigating the effect of PSE on other perception- or behaviour-based variables of accounting system use. In our research we address Dunn and Grabski’s suggestion. We set out to develop and evaluate a multi-item measurement instrument for PSE. We also plan to use this instrument in an experimental evaluation of the REA model, looking for other benefits that those found by Dunn and Grabski. In this paper we describe the item generation and item refinement steps in the measure development process. We describe how we started from a conceptual definition of PSE using a linguistics- and semiotics-based theoretical framework for understanding quality in conceptual modeling that was found in the literature. Next, we proposed seven measurement instrument items corresponding to well-known quality properties for conceptual schemas that we first mapped onto the PSE construct. As a pre-test, our initial PSE measurement instrument was applied in a laboratory experiment, allowing us to assess the validity and reliability of the measure. In particular we investigated whether our instrument measures another construct than those measured by user information satisfaction (UIS) and perceived ease of use (PEOU) measures. Based on this analysis, the measurement instrument was further refined to obtain acceptable levels of validity and reliability.