acm - an acm publication

Articles

Empirical software research
an interview with Dag Sjøberg, University of Oslo, Norway

Ubiquity, Volume 2011 Issue June, June 2011 | BY Walter Tichy 

|

Full citation in the ACM Digital Library  | PDF


Connection Failure

COMMENTS

As a long-term colleague, and having worked in, for, and with the industry my whole career, I commend the viewpoints expressed. I think we agree that developing scientific knowledge to support decision makers requires high-quality empirical studies, synthesis of empirical evidence, and development of empirical theories. A particular challenge as I see it, therefore, is to do empirical research that is interesting and that can have an impact. Its fairly easy to describe what people do and say, or how systems behave and change through experiments or observational methods, but to go beyond that and suggest insights and explanations that offer a more theoretical understanding is not so easy. This makes it even more important, as you say, to spend our limited resources on the fundamental issues. These issues are not necessarily revealed in the common strategy of looking for gaps in the literature to identify issues for future research. I think we should rather question the underlying assumptions of much of the existing literature and rather dig deeper into fundamental questions such as How to design complex software systems in particular environments? and What ways of coordinating global software teams are likely to facilitates efficient development? In my opinion, the only way of providing deeper insights on such questions is to focus on a fruitful interplay between empirical studies and theory development.

��� Tore Dybå, Mon, 15 Aug 2011 23:34:27 UTC

Many thanks for your encouragement and useful comments. I certainly agree on your three directions for future effort. One great challenge, though, is to attract sufficient funding. It seems very difficult to convince funding bodies to support long-term research programmes and projects in computing on fundamental issues. I believe we have a job to do to improve the understanding of computer science in general and the empirical component in particular. For example, at present, much of the money from the bodies that fund computer science research goes to research on the *use* of social media. Furthermore, I frequently meet the attitude that long-term research in computing is not worthwhile because much of the necessary research is done by the IT industry anyway and long-term projects dont pay off because the research quickly becomes obsolete anyway. To investigate this last claim, I actually measured the citation half-life (an indicator of how long the research lives) of 8300 journals in ISI Web of Knowledge in all disciplines. It turned out that the research front moves quickest in Immunology, and Molecular Biology and Genetics, and slowest in Economics and Business, and Mathematics. Surprisingly, Computer Science is in the middle. (The results were published in CACM September 2010.) Anyway, as you state, to make real progress the international community needs to join forces to establish larger research programmes on the fundamental issues, with a strong empirical component, similar to what we find in other disciplines such as physics and medicine.

��� Dag Sjoberg, Tue, 05 Jul 2011 22:19:23 UTC

Regarding the Sjoberg interviw. This interview contains many excellent ideas regarding the value and the approaches to be taken for empirical studies. I agree with everything that is being said. Nevertheless, I like to encourage you and your co-workers to do more work in at least three directions: (1)Better identify those (middle range) theories that help explain why the things you observe happen, and do not restrict yourself to mathematics. As you say, correlations can lie. They even may defy causality. Only a feasible theory can help to avoid wrong conclusions. (2)Do not shrink away from what other people call the fundamental questions of software engineering. If this relates to the usability and value of software, only empirical studies will help. If they are concerned about the development process and methods, there are no alternatives either. (3)As you have now crossed the boundary between inexpensive and costly studies, the more necessary it becomes to coordinate your efforts within the international community. But do not let yourself be slowed down.

��� Al Endres, Stuttgart, Germany, Mon, 27 Jun 2011 17:26:26 UTC

POST A COMMENT
Leave this field empty