Asymmetry in Identification of Multiplicity Errors

Company policies and business rules can be represented by multiplicities in a UML class diagram. Mistakes made in the multiplicities may result in a poorly designed database that does not faithfully represent the enterprise’s business operations and impedes compliance with its policies. System auditors need to be able to validate an enterprise’s conceptual data model, including its multiplicities, against the underlying reality. Prior research has revealed conflicting asymmetries in accuracy for minimum multiplicity error identification tasks. Two experiments were conducted to gain a further understanding and help resolve this apparent contradiction, and to explore whether any asymmetry exists in the identification of errors in maximum multiplicities. Results indicate the minimum multiplicity asymmetry in prior studies with system evaluators that was contrary to ontology research was likely an artifact of the task prompts. Without those prompts accuracy is consistent with the ontology research stream. Accuracy in evaluating minimum multiplicities in both experiments was greatest when the underlying semantics represented mandatory participation. Results also indicate an asymmetry for evaluation of maximum multiplicities such that accuracy is greatest when the underlying semantics represented flexible participation.