Academic solutions do not work in the real world. On the day I am writing this abstract I find professor @hemmatihadi publicly bemoaning his struggles "..applying several fully automated specification mining [approaches] in an industrial setting.. [because] none worked as expected." Let me emphasize his point: not even a single academic approach worked in this real world setting. To those working at the boundary of research and application this is, unfortunately, not surprising.
Yet why should these (and many other) academic solutions fail when put to the test? They were created by some of the brightest minds in our field. I believe the fault is not in these researchers. I believe that these solutions fail because they were developed and evaluated in an artificial context. While data analysis and lab experiments are crucial to scientific progress, embodying an idea as a usable product allows it to be poked, prodded, used, and mis-used just like it would in the real world. For this, there is no substitute, which is why I conduct the majority of my research as what I am calling *Product-Based Research*.
This talk will focus on how to conduct Product-Based Research. Using three of my recent projects as examples, including the FlowLight project which was featured by major media outlets (e.g., BBC Radio, Swedish Radio, and The Wall Street Journal), I will discuss the advantages of this approach. I will also discuss the sobering costs. By the end of this talk you will be able to use this approach, when appropriate, in your own work.
Editor-in-Chief, Journal of Systems and Software
Software Engineering Research Group, ABB Corporate Research, USA
David C. Shepherd is a Senior Principal Scientist with ABB Corporate Research where he leads a group focused on improving developer productivity and increasing software quality. His background, including becoming employee number nine at a successful software tools spinoff and working extensively on popular open source projects, has focused his research on bridging the gap between academic ideas and viable industrial tools. His main research interests to date have centered on software tools that improve developers search and navigation behavior.
