Understanding user errors in database query.

Abstract

SQL (Structured Query Language) is the de facto standard language for querying databases. Unfortunately, users commit many errors when using SQL. The most common error is the omission of the "join statement," which indicates how to combine multiple database tables. This error wastes users' time for error correction, at an annual cost of approximately $53 million, but also returns answers that may be undetectably wrong. This dissertation guides the solution of this serious user error. This error can be fixed in many ways. In order to choose the most effective way, we need to consider the cognitive causes of the error, the set of available solutions, and the costs to the individuals involved. This dissertation first attempts to narrow the solution space by isolating the effective cognitive causes of this error. It then evaluates the remaining solutions for their costs and benefits. Isolation of the cognitive causes required construction of a new model of query writing, combining a task analysis with known characteristics of the human information processor. Because an analysis of existing data was not sufficient to isolate the causes of the error, a controlled study was run. Results of that study indicate that four factors contributed to the join omission: incorrect user knowledge, absence of a cue, overloading the user's working memory capacity, and functional fixedness. Since many factors contributed to the error, the multiple, appropriate solutions were evaluated for their costs and benefits to the end user, the software designer, and the information systems manager. This evaluation indicates that end users will always prefer an intelligent interface, since it requires the least time, but information systems managers will choose different solutions depending on the number of query users in their organization. This dissertation contributes a cognitive model of query writing, used to isolate causes of user-interface errors, and a method for evaluating the costs and benefits of various solutions to user-interface problems.