Empirical Paradigm - The Role of Experiments

This article discusses the role of formal experiments in empirical software engineering. I take the view that the role of experiments has been overemphasised. Laboratory experiments are not representative of industrial software engineering tasks, so do not provide us with a reliable assessment of the effect of our techniques and tools. I suggest we need to concentrate a larger proportion of our research effort on industrial quasi-experiments and case studies. Methodologies for these empirical methods are well-understood in the social science and would appear to be appropriate mechanisms for investigating many software engineering research questions. In addition, I believe we need to make the results of empirical software engineering more visible and relevant to practitioners. To influence practitioners I suggest that we need to produce evidence-based text books and evidence-based software engineering standards.