The classical scientific method has been settled through the last centuries as a cyclic, iterative process of observation, hypothesis formulation, and confirmation/refutation of hypothesis through experimentation. This “experimental scientific method” was mainly developed in the context of natural sciences dealing with the physical world, such as Mechanics, Thermodynamics, Electromagnetism, Chemistry, and so on. But when trying to apply this classical view of the scientific method to the various branches of Computer Science and Computer Engineering, among which Software Engineering, there are two kinds of obstacles. First, Computer Science is rooted both in formal sciences such as Mathematics and experimental sciences such as Physics, and therefore, an excessive emphasis on the experimental side is not appropriate to give a full account of this kind of scientific activity. Second, the production of software systems has to deal not only with the behavior of complex physical systems such as computers, but also with the behavior of complex human systems (developers interacting with stakeholders, for instance, or users interacting with machines) where educational, cultural, sociological, and economical factors are essential. Therefore, empirical methods in their narrow sense, even though valuable in some respects, are rather limited to understand a reality that exceeds the mere physical world. Moreover, neither formal nor empirical methods can provide a full account of scientific activity, which relies on something that is beyond any established method. Qualitative (i.e. meta-methodical) reasoning plays the directive role in scientific activity. In this chapter, the authors claim that acknowledging a plurality of research methods in software engineering will benefit the advancement of this branch of science.
Selecting Empirical Methods for Software Engineering Research
