ארנון שטורם

Model-based software engineering places models in the center of the development process. To support this notion, multiple modeling languages are available, and the visual ones are widely used. Nevertheless, visual modeling languages are limited in their expressiveness and sometime might introduce ambiguity into the models. To overcome these limitations, model-based constraint languages have emerged, yet, their usage is limited, probably due to a misconception that they are difficult to work with. In this paper, we challenge this misconception by comparing the use of three constraint languages: OCL, Java, and natural language in understanding and developing model-based constraints. The comparison was made through a controlled experiment with 68 information systems engineering undergraduate students. We found out that using natural language results in shorter times to perform the tasks. Yet, using OCL results in increased accuracy for specifying new constraints, in particular, in complex settings.