Evidence-based Practices Informing Computing (EPIC)

The use of evidence-based practices in research provides a means of aggregating the outcomes from a set of (primary) empirical studies, in order to obtain a 'bigger picture' of a phenomenon of interest. The goal for evidence-based research is to ensure that practices and standards that are based on the outcomes from evidence-based practices such as Systematic Literature Review (often termed 'secondary studies') will be as objective and unbiased as possible, and should not be unduly influenced by the results from individual studies.This approach is now used extensively in clinical medicine, as well as having been adopted for use in such diverse domains as education, nursing & midwifery, librarianship, and environmental science. We ourselves have been investigating how the approach may be adapted to the characteristics of software engineering--where empirical studies are fewer in number, complicated by individual skill levels of those involved, and rarely well-indexed, making it difficult to find them on a systematic basis. Despite these factors, we (and others) have demonstrated that it is quite practical to employ evidence-based practices in software engineering, and the present proposal is intended to move beyond that initial investigation to address such questions as where they might be most effectively deployed, what limitations upon the outcomes of studies might arise from the characteristics of the domain, and how researchers, practitioners and decision-makes might be persuaded (with evidence!) of the value of this approach?To do this, we propose to adopt a multiple-case, embedded form of case study for our research methodology, drawing together the outcomes of several different case studies. In addition, this will be embedded in a participant-observer framework, since it is impractical for us to be external observers. By employing such an approach to the use of case studies, we expect to be able to determine how effective the use of evidence-based practices are likely to be as a means of delivering solid evidence to software engineers, managers and policy-makers.