Handling of elicited information in requirements engineering: good practices

Abstract

Capturing information in the requirements process is usually done through interviews. However, despite its frequent use, the practical indications available on how to carry interviews out and how to subsequently use the information to elaborate models that reflect the knowledge acquired, do not prevent the appearance of some relevant systematic errors. Both in the literature and the experiments carried out, it has been detected that the interviewer and/or the modeler usually introduce subjective information not provided by the information source, but coming from the interviewer or modeler’s prior knowledge. This distortion in the modeled information is often transferred to other models and software artifacts, being very later detected. This led to carrying out an experiment that studied in detail how the information was elicited through an interview and how it was processed to produce a model written in natural language. The use of incorrectly named concepts, non-existent concepts, misunderstood information, and the induction of vocabulary by the interviewer were detected in the model. Based on the distortions found in the modeled information, preliminary recommendations were made about how to conduct interviews and how to process the information for modeling.