| Businesses rely extensively on data resources stored in database management systems documented using graphical data representations such as Entity-Relationship diagrams. Users range from trained information systems professionals to knowledge workers with little training in database technology. Increasingly, these users are given direct access to organizational data to support their decision-making processes. They formulate ad hoc queries using a database manipulation language (textual or graphical) often assisted by a graphical database representation.; There are many approaches to developing such representations. Some manifest fundamental, ontological differences in the way they represent the reality recorded in the database. Some models represent temporal dynamics by focusing on things and states while others focus on events that cause the state of things to change. This research draws on cognitive science literature to posit that users of event-driven graphical data representations will formulate ad hoc queries more quickly, more accurately, and have higher confidence in the correctness of their queries than will users of state driven representations.; An experiment was conducted involving 342 students from six universities in the United States and Europe. Administered over the Internet, this experiment required subjects to review a textual description of a business process and formulate queries in Structured Query Language to answer questions about the business activity of a company with which they were not familiar. Subjects were randomly assigned to one of three graphical representations (Entity-Relationship diagrams) to support their participation: an event-driven representation and two state driven representations. The experimental instrument collected various performance measures as subjects completed the experiment.; The results indicate that the use of event-driven models allows users to better predict the accuracy of their queries. Partial support was found to suggest that users of event-driven models formulate queries more quickly than users of state-driven models. Both of these findings are limited to subjects who expressed low and very low comfort reading Entity-Relationship models. The results indicated no significant difference among treatments with respect to query correctness, confidence, number of queries tested, time spent viewing supporting documentation, number of moves to solve the task, or length of time viewing the results of queries tested. |
