Communicating the Message: Translating Tasks into Queries in a Database ContextIEEE Transactions on Professional Communication Srinivasan, A. and Irwin, G.
Authors from the College of Business:
Gretchen Irwin, Associate Professor, Ehrhardt Keefe Steiner and Hottman Research Fellow
Knowledge workers often interact with databases to accomplish their tasks. For example, sales managers may query a database to see how sales have varied over time by product and location. To effectively retrieve data from a database, the user needs to understand: 1) what information is in the database (the data model), and 2) how to pull information out of the database (the query language). Most business database systems use a data model and query language based on constructs such as tables, primary keys, and foreign keys. To formulate a query in this environment, the sales manager may have to know that sales information is spread across several tables that are linked together by certain foreign keys. We consider these constructs to be fairly low-level, technical, and “cognitively distant” from the higher-level constructs relevant to the user (e.g., sales, products, and locations). The user must spend significant cognitive effort to translate his task into the lower-level constructs required by the query language.
Our research examines whether a data model and query language with higher-level constructs can facilitate query formulation. For example, if the user can query a “sales” object as if it were one thing containing product and location information, would the query be easier to write and have fewer errors? We conducted an in-depth verbal protocol study, where users “thought aloud” as they wrote queries for tasks of increasing complexity. Some users worked with higher-level constructs (e.g., a single, complex sales object) and some worked with lower-level constructs (e.g., a sale table, sale detail table, product table, and store table all linked by foreign keys). We found that as the tasks became more complex, the queries written with higher-level constructs were more correct than the ones written with lower-level constructs. We discuss this finding in the context of prior research, which has shown mixed results for the benefits of using higher-level constructs, and propose avenues for further research.