Modeling as Constrained Problem Solving: An Empirical Study of the Data Modeling Process

Abstract

Modeling is a powerful tool for managing complexity in problem solving. Problem solvers usually build a simplified model of the real world and then use it to generate a solution to their problem. To date, however, little is known about how people actually behave when building a model. This study concentrates on data modeling, which involves the representation of different types of data and their interrelationships. It reports on two laboratory studies, in which subjects engage in data modeling to solve a complex problem. Using a think-aloud process-tracing methodology, we examine the data modeling behavior as a set of activities that are managed by several heuristics. We found that some heuristics were effective in reducing the complexity of the problem. An important aspect we observed was how subjects moved across levels of abstraction in the problem representation. Overall, these observations help to explain how people deal with complexity in data modeling. They also suggest that it may be advantageous to design systems that support work at various levels of abstraction and support transitions among those levels.