A family of experiments to generate graphical user interfaces from BPMN models with stereotypes

Context: A significant gap separates Business Process Model and Notation (BPMN) models representing processes from the design of Graphical User Interfaces (GUIs). Objective: This paper reports on a family of experiments to validate a method to automatically generate GUIs from BPMN models using stereotypes complemented with UML class primitives, and transformation rules. Method: We conducted two replications (23 and 31 subjects respectively) in which we compared two methods to generate GUIs from BPMN models; one automatic (using Stereotyped BPMN models) and one manual (using Non-stereotyped BPMN models). The study focuses on comparing effort, accuracy, and satisfaction (in terms of perceived ease of use (PEOU), perceived usefulness (PU), and intention to use (ITU)). Results: Results yield significant differences for Effort, Accuracy, and ITU. Effort is lower for the Non-stereotyped method, while accuracy and ITU are higher for the Stereotyped one. If we consider only experimental units whose BPMN models show an accuracy over 75% compared to those of the experimenters’ solution, the difference in accuracy for the designed GUIs is even more significant; in contrast, differences for effort and ITU are reduced. Conclusions: The use of the Stereotyped method reduces the possibility of errors in the process of designing GUIs.